test(ci): read sparse android guard files from git

Merged via squash.

Prepared head SHA: 562f2ac5a8
Co-authored-by: jesse-merhi <79823012+jesse-merhi@users.noreply.github.com>
Reviewed-by: @jesse-merhi

.agents/skills/claw-score/SKILL.md

@@ -146,7 +146,7 @@ Default guidance:
 
 Default Completeness bands:
 
-- `Lovable` (95-100): complete across expected workflows, variants, and
+- `Clawesome` (95-100): complete across expected workflows, variants, and
   recovery branches, with only minor polish gaps.
 - `Stable` (80-95): the expected workflow set is broadly present, with only
   bounded missing branches.
@@ -172,7 +172,7 @@ Default Completeness bands:
 
 Bands:
 
-- `Lovable`: 95-100
+- `Clawesome`: 95-100
 - `Stable`: 80-95
 - `Beta`: 70-80
 - `Alpha`: 50-70

.github/codex/prompts/maturity-scorecard-agent.md

@@ -0,0 +1,23 @@
+# OpenClaw Maturity Scorecard Agent
+
+You are refreshing the OpenClaw maturity score source for a release scorecard.
+
+Goal: use the `$claw-score` skill to refresh `qa/maturity-scores.yaml` for every active surface in `taxonomy.yaml`, using the current repository and the release evidence artifacts in `.artifacts/maturity-evidence`.
+
+Allowed tracked paths:
+
+- `qa/maturity-scores.yaml`
+
+Hard limits:
+
+- Do not edit generated docs, taxonomy, workflows, scripts, package metadata, lockfiles, tests, or application code.
+- Do not render docs. The workflow renders docs after validating the score source.
+- Keep the score source schema valid for QA Lab maturity score validation.
+
+Required workflow:
+
+1. Use the `$claw-score` skill before editing.
+2. Read `taxonomy.yaml`, any existing maturity score file, and the release evidence artifacts.
+3. Refresh scores for every active surface in `taxonomy.yaml`.
+4. Run the QA Lab maturity score validation used by this repository.
+5. If no defensible score update is possible, leave a valid `qa/maturity-scores.yaml` and explain the uncertainty in the final message.

.github/workflows/crabbox-hydrate.yml

@@ -490,7 +490,7 @@ jobs:
           if (-not $env:CRABBOX_ID -or $env:CRABBOX_ID -notmatch '^[A-Za-z0-9._-]+$') {
             Write-Error "Invalid crabbox_id"
           }
-          $actionsRoot = Join-Path $HOME ".crabbox\actions"
+          $actionsRoot = "C:\ProgramData\crabbox\actions"
           New-Item -ItemType Directory -Force $actionsRoot | Out-Null
           $state = Join-Path $actionsRoot "$env:CRABBOX_ID.env"
           $envFile = Join-Path $actionsRoot "$env:CRABBOX_ID.env.ps1"
@@ -546,7 +546,7 @@ jobs:
           if ($env:CRABBOX_KEEP_ALIVE_MINUTES -match '^[0-9]+$') {
             $minutes = [int]$env:CRABBOX_KEEP_ALIVE_MINUTES
           }
-          $stop = Join-Path $HOME ".crabbox\actions\$env:CRABBOX_ID.stop"
+          $stop = Join-Path "C:\ProgramData\crabbox\actions" "$env:CRABBOX_ID.stop"
           $deadline = (Get-Date).AddMinutes($minutes)
           while ((Get-Date) -lt $deadline) {
             if (Test-Path $stop) {

.github/workflows/maturity-scorecard.yml

@@ -12,6 +12,40 @@ on:
         required: true
         default: main
         type: string
+      expected_sha:
+        description: Optional full SHA that ref must resolve to
+        required: false
+        default: ""
+        type: string
+  workflow_call:
+    inputs:
+      qa_evidence_run_id:
+        description: Optional workflow run id containing qa-evidence.json
+        required: false
+        default: ""
+        type: string
+      ref:
+        description: OpenClaw branch, tag, or SHA containing the maturity score source
+        required: true
+        type: string
+      expected_sha:
+        description: Optional full SHA that ref must resolve to
+        required: false
+        default: ""
+        type: string
+    secrets:
+      OPENAI_API_KEY:
+        description: OpenAI API key used by live QA profile scenarios
+        required: true
+      OPENCLAW_MATURITY_SCORECARD_AGENT_OPENAI_API_KEY:
+        description: Optional OpenAI API key used by maturity scorecard agent steps
+        required: false
+      GH_APP_PRIVATE_KEY:
+        description: Optional GitHub App private key for generated docs PR creation
+        required: false
+      GH_APP_PRIVATE_KEY_FALLBACK:
+        description: Optional fallback GitHub App private key for generated docs PR creation
+        required: false
 
 permissions:
   actions: read
@@ -43,14 +77,25 @@ jobs:
       - name: Validate selected ref
         id: validate
         env:
+          EXPECTED_SHA: ${{ inputs.expected_sha }}
           INPUT_REF: ${{ inputs.ref }}
         shell: bash
         run: |
           set -euo pipefail
 
           selected_revision="$(git rev-parse HEAD)"
+          expected_sha="${EXPECTED_SHA,,}"
           trusted_reason=""
 
+          if [[ -n "${expected_sha// }" && ! "$expected_sha" =~ ^[0-9a-f]{40}$ ]]; then
+            echo "expected_sha must be a full 40-character SHA; got: ${EXPECTED_SHA}" >&2
+            exit 1
+          fi
+          if [[ -n "${expected_sha// }" && "${selected_revision,,}" != "$expected_sha" ]]; then
+            echo "Ref '${INPUT_REF}' resolved to ${selected_revision}, expected ${EXPECTED_SHA}." >&2
+            exit 1
+          fi
+
           git fetch --no-tags origin +refs/heads/main:refs/remotes/origin/main
 
           if git merge-base --is-ancestor "$selected_revision" refs/remotes/origin/main; then
@@ -87,8 +132,9 @@ jobs:
     if: ${{ inputs.qa_evidence_run_id == '' }}
     uses: ./.github/workflows/qa-profile-evidence.yml
     with:
-      ref: ${{ needs.validate_selected_ref.outputs.selected_revision }}
-      qa_profile: all
+      ref: ${{ inputs.ref }}
+      expected_sha: ${{ needs.validate_selected_ref.outputs.selected_revision }}
+      qa_profile: release
     secrets:
       OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
 
@@ -192,8 +238,8 @@ jobs:
           }
 
           const evidence = JSON.parse(fs.readFileSync(evidencePath, "utf8"));
-          if (evidence.profile !== "all") {
-            throw new Error(`qa-evidence.json profile must be all, got ${JSON.stringify(evidence.profile)}`);
+          if (evidence.profile !== "release") {
+            throw new Error(`qa-evidence.json profile must be release, got ${JSON.stringify(evidence.profile)}`);
           }
 
           const artifactDir = path.dirname(evidencePath);
@@ -210,14 +256,75 @@ jobs:
           const manifestPath = path.join(artifactDir, manifestNames[0]);
           const manifest = JSON.parse(fs.readFileSync(manifestPath, "utf8"));
           const manifestProfile = manifest.qaProfile ?? evidence.profile;
-          if (manifestProfile !== "all") {
-            throw new Error(`QA evidence manifest profile must be all, got ${JSON.stringify(manifestProfile)}`);
+          if (manifestProfile !== "release") {
+            throw new Error(`QA evidence manifest profile must be release, got ${JSON.stringify(manifestProfile)}`);
           }
           if (manifest.targetSha !== targetSha) {
             throw new Error(`QA evidence manifest targetSha ${manifest.targetSha} does not match selected ref ${targetSha}`);
           }
           NODE
 
+      - name: Ensure maturity scorecard agent key exists
+        env:
+          OPENAI_API_KEY: ${{ secrets.OPENCLAW_MATURITY_SCORECARD_AGENT_OPENAI_API_KEY || secrets.OPENAI_API_KEY }}
+        run: |
+          set -euo pipefail
+          if [[ -z "${OPENAI_API_KEY:-}" ]]; then
+            echo "Missing OPENCLAW_MATURITY_SCORECARD_AGENT_OPENAI_API_KEY or OPENAI_API_KEY secret." >&2
+            exit 1
+          fi
+
+      - name: Run Codex maturity scorecard agent
+        uses: openai/codex-action@e0fdf01220eb9a88167c4898839d273e3f2609d1
+        env:
+          MATURITY_EVIDENCE_DIR: .artifacts/maturity-evidence
+          MATURITY_SCORES_PATH: qa/maturity-scores.yaml
+          MATURITY_TAXONOMY_PATH: taxonomy.yaml
+        with:
+          openai-api-key: ${{ secrets.OPENCLAW_MATURITY_SCORECARD_AGENT_OPENAI_API_KEY || secrets.OPENAI_API_KEY }}
+          prompt-file: .github/codex/prompts/maturity-scorecard-agent.md
+          model: ${{ vars.OPENCLAW_CI_OPENAI_MODEL_BARE }}
+          effort: high
+          sandbox: workspace-write
+          safety-strategy: drop-sudo
+
+      - name: Enforce focused maturity score patch
+        run: |
+          set -euo pipefail
+          git restore --staged :/
+
+          allowed='^qa/maturity-scores\.yaml$'
+          bad_tracked="$(
+            git diff --name-only HEAD -- | while IFS= read -r path; do
+              if [[ ! "$path" =~ $allowed ]]; then
+                printf '%s\n' "$path"
+              fi
+            done
+          )"
+          if [[ -n "$bad_tracked" ]]; then
+            echo "Maturity scorecard agent touched forbidden tracked paths:"
+            printf '%s\n' "$bad_tracked"
+            exit 1
+          fi
+
+          bad_untracked="$(
+            git ls-files --others --exclude-standard | while IFS= read -r path; do
+              if [[ "$path" != "qa/maturity-scores.yaml" ]]; then
+                printf '%s\n' "$path"
+              fi
+            done
+          )"
+          if [[ -n "$bad_untracked" ]]; then
+            echo "Maturity scorecard agent created forbidden untracked paths:"
+            printf '%s\n' "$bad_untracked"
+            exit 1
+          fi
+
+          if [[ ! -f qa/maturity-scores.yaml ]]; then
+            echo "Maturity scorecard agent must produce qa/maturity-scores.yaml." >&2
+            exit 1
+          fi
+
       - name: Validate maturity score sources
         run: |
           node --import tsx --input-type=module <<'NODE'
@@ -260,6 +367,7 @@ jobs:
             --strict-inputs
 
       - name: Create generated docs PR app token
+        if: ${{ github.event_name == 'workflow_dispatch' }}
         id: app-token
         continue-on-error: true
         uses: actions/create-github-app-token@bcd2ba49218906704ab6c1aa796996da409d3eb1 # v3
@@ -270,7 +378,7 @@ jobs:
           permission-pull-requests: write
 
       - name: Create generated docs PR fallback app token
-        if: ${{ steps.app-token.outcome == 'failure' }}
+        if: ${{ github.event_name == 'workflow_dispatch' && steps.app-token.outcome == 'failure' }}
         id: app-token-fallback
         uses: actions/create-github-app-token@bcd2ba49218906704ab6c1aa796996da409d3eb1 # v3
         with:
@@ -280,6 +388,7 @@ jobs:
           permission-pull-requests: write
 
       - name: Open generated docs PR
+        if: ${{ github.event_name == 'workflow_dispatch' }}
         env:
           GH_TOKEN: ${{ steps.app-token.outputs.token || steps.app-token-fallback.outputs.token }}
           QA_EVIDENCE_RUN_ID: ${{ inputs.qa_evidence_run_id }}
@@ -291,7 +400,7 @@ jobs:
             exit 1
           fi
 
-          if [[ -z "$(git status --porcelain -- qa/maturity-scores.yaml docs/maturity-scores.yaml docs/maturity/scorecard.md docs/maturity/taxonomy.md)" ]]; then
+          if [[ -z "$(git status --porcelain -- qa/maturity-scores.yaml docs/maturity/scorecard.md docs/maturity/taxonomy.md)" ]]; then
             {
               echo
               echo "- Pull request: skipped; generated scorecard matches selected ref"
@@ -311,9 +420,6 @@ jobs:
           git fetch --no-tags --depth=1 origin "refs/heads/${branch}:refs/remotes/origin/${branch}" || true
           git switch -C "$branch"
           git add qa/maturity-scores.yaml docs/maturity/scorecard.md docs/maturity/taxonomy.md
-          if git ls-files --error-unmatch docs/maturity-scores.yaml >/dev/null 2>&1 || [[ -e docs/maturity-scores.yaml ]]; then
-            git add docs/maturity-scores.yaml
-          fi
           git commit -m "docs: update maturity scorecard"
           git push --force-with-lease origin "$branch"
 

.github/workflows/openclaw-release-checks.yml

@@ -44,6 +44,11 @@ on:
         required: false
         default: false
         type: boolean
+      run_maturity_scorecard:
+        description: Render advisory maturity scorecard release docs; default release checks rely on dedicated package, QA, live, and E2E gates
+        required: false
+        default: false
+        type: boolean
       rerun_group:
         description: Release check group to run
         required: false
@@ -106,6 +111,7 @@ jobs:
       mode: ${{ steps.inputs.outputs.mode }}
       release_profile: ${{ steps.inputs.outputs.release_profile }}
       run_release_soak: ${{ steps.inputs.outputs.run_release_soak }}
+      run_maturity_scorecard: ${{ steps.inputs.outputs.run_maturity_scorecard }}
       rerun_group: ${{ steps.inputs.outputs.rerun_group }}
       live_suite_filter: ${{ steps.inputs.outputs.live_suite_filter }}
       cross_os_suite_filter: ${{ steps.inputs.outputs.cross_os_suite_filter }}
@@ -279,6 +285,7 @@ jobs:
           RELEASE_MODE_INPUT: ${{ inputs.mode }}
           RELEASE_PROFILE_INPUT: ${{ inputs.release_profile }}
           RELEASE_RUN_RELEASE_SOAK_INPUT: ${{ inputs.run_release_soak }}
+          RELEASE_RUN_MATURITY_SCORECARD_INPUT: ${{ inputs.run_maturity_scorecard }}
           RELEASE_RERUN_GROUP_INPUT: ${{ inputs.rerun_group }}
           RELEASE_LIVE_SUITE_FILTER_INPUT: ${{ inputs.live_suite_filter }}
           RELEASE_CROSS_OS_SUITE_FILTER_INPUT: ${{ inputs.cross_os_suite_filter }}
@@ -319,6 +326,12 @@ jobs:
           else
             run_release_soak=true
           fi
+          run_maturity_scorecard="$(printf '%s' "$RELEASE_RUN_MATURITY_SCORECARD_INPUT" | tr '[:upper:]' '[:lower:]')"
+          if [[ "$run_maturity_scorecard" != "true" && "$run_maturity_scorecard" != "1" && "$run_maturity_scorecard" != "yes" ]]; then
+            run_maturity_scorecard=false
+          else
+            run_maturity_scorecard=true
+          fi
           release_profile="$RELEASE_PROFILE_INPUT"
           if [[ "$release_profile" == "minimum" ]]; then
             release_profile=beta
@@ -422,6 +435,7 @@ jobs:
             printf 'mode=%s\n' "$RELEASE_MODE_INPUT"
             printf 'release_profile=%s\n' "$release_profile"
             printf 'run_release_soak=%s\n' "$run_release_soak"
+            printf 'run_maturity_scorecard=%s\n' "$run_maturity_scorecard"
             printf 'rerun_group=%s\n' "$RELEASE_RERUN_GROUP_INPUT"
             printf 'live_suite_filter=%s\n' "$RELEASE_LIVE_SUITE_FILTER_INPUT"
             printf 'cross_os_suite_filter=%s\n' "$RELEASE_CROSS_OS_SUITE_FILTER_INPUT"
@@ -444,6 +458,7 @@ jobs:
           RELEASE_MODE: ${{ inputs.mode }}
           RELEASE_PROFILE: ${{ steps.inputs.outputs.release_profile }}
           RUN_RELEASE_SOAK: ${{ steps.inputs.outputs.run_release_soak }}
+          RUN_MATURITY_SCORECARD: ${{ steps.inputs.outputs.run_maturity_scorecard }}
           RELEASE_RERUN_GROUP: ${{ inputs.rerun_group }}
           RELEASE_LIVE_SUITE_FILTER: ${{ inputs.live_suite_filter }}
           RELEASE_CROSS_OS_SUITE_FILTER: ${{ inputs.cross_os_suite_filter }}
@@ -461,6 +476,7 @@ jobs:
             echo "- Cross-OS mode: \`${RELEASE_MODE}\`"
             echo "- Release profile: \`${RELEASE_PROFILE}\`"
             echo "- Release soak lanes: \`${RUN_RELEASE_SOAK}\`"
+            echo "- Maturity scorecard docs: \`${RUN_MATURITY_SCORECARD}\`"
             echo "- Rerun group: \`${RELEASE_RERUN_GROUP}\`"
             if [[ -n "${RELEASE_LIVE_SUITE_FILTER// }" ]]; then
               echo "- Live suite filter: \`${RELEASE_LIVE_SUITE_FILTER}\`"
@@ -767,6 +783,20 @@ jobs:
       OPENCLAW_QA_CONVEX_SITE_URL: ${{ secrets.OPENCLAW_QA_CONVEX_SITE_URL }}
       OPENCLAW_QA_CONVEX_SECRET_CI: ${{ secrets.OPENCLAW_QA_CONVEX_SECRET_CI }}
 
+  maturity_scorecard_release_checks:
+    name: Render maturity scorecard release docs
+    needs: [resolve_target]
+    if: contains(fromJSON('["all","qa"]'), needs.resolve_target.outputs.rerun_group) && needs.resolve_target.outputs.run_maturity_scorecard == 'true'
+    permissions:
+      actions: read
+      contents: read
+    uses: ./.github/workflows/maturity-scorecard.yml
+    with:
+      ref: ${{ needs.resolve_target.outputs.ref }}
+      expected_sha: ${{ needs.resolve_target.outputs.revision }}
+    secrets:
+      OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
+
   qa_lab_parity_lane_release_checks:
     name: Run QA Lab parity lane (${{ matrix.lane }})
     needs: [resolve_target]
@@ -853,7 +883,7 @@ jobs:
           name: release-qa-parity-${{ matrix.lane }}-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
       - name: Record advisory status
         if: always()
@@ -959,7 +989,7 @@ jobs:
           name: release-qa-parity-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
       - name: Record advisory status
         if: always()
@@ -1131,7 +1161,7 @@ jobs:
           name: release-qa-runtime-parity-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
       - name: Record advisory status
         if: always()
@@ -1241,13 +1271,13 @@ jobs:
             --output .artifacts/qa-e2e/runtime-parity-standard-report/qa-runtime-tool-coverage-report.md
 
       - name: Upload runtime tool coverage artifacts
-        if: always()
+        if: ${{ always() && steps.verify_runtime_parity_status.outputs.ready == 'true' }}
         uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
         with:
           name: release-qa-runtime-tool-coverage-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/runtime-parity-standard-report/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
   qa_live_matrix_release_checks:
     name: Run QA Lab live Matrix lane
@@ -1327,7 +1357,7 @@ jobs:
           name: release-qa-live-matrix-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
       - name: Record advisory status
         if: always()
@@ -1467,7 +1497,7 @@ jobs:
           name: release-qa-live-telegram-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
       - name: Record advisory status
         if: always()
@@ -1607,7 +1637,7 @@ jobs:
           name: release-qa-live-discord-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
       - name: Record advisory status
         if: always()
@@ -1750,7 +1780,7 @@ jobs:
           name: release-qa-live-whatsapp-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
       - name: Record advisory status
         if: always()
@@ -1890,7 +1920,7 @@ jobs:
           name: release-qa-live-slack-${{ needs.resolve_target.outputs.revision }}
           path: .artifacts/qa-e2e/
           retention-days: 14
-          if-no-files-found: warn
+          if-no-files-found: error
 
       - name: Record advisory status
         if: always()
@@ -1946,6 +1976,7 @@ jobs:
       - docker_e2e_release_checks
       - package_acceptance_release_checks
       - qa_lab_parity_lane_release_checks
+      - maturity_scorecard_release_checks
       - qa_lab_parity_report_release_checks
       - qa_lab_runtime_parity_release_checks
       - runtime_tool_coverage_release_checks
@@ -2031,6 +2062,7 @@ jobs:
             "docker_e2e_release_checks=${{ needs.docker_e2e_release_checks.result }}" \
             "package_acceptance_release_checks=${{ needs.package_acceptance_release_checks.result }}" \
             "qa_lab_parity_lane_release_checks=${{ needs.qa_lab_parity_lane_release_checks.result }}" \
+            "maturity_scorecard_release_checks=${{ needs.maturity_scorecard_release_checks.result }}" \
             "qa_lab_parity_report_release_checks=${{ needs.qa_lab_parity_report_release_checks.result }}" \
             "qa_lab_runtime_parity_release_checks=${{ needs.qa_lab_runtime_parity_release_checks.result }}" \
             "runtime_tool_coverage_release_checks=${{ needs.runtime_tool_coverage_release_checks.result }}" \

.github/workflows/openclaw-release-publish.yml

@@ -1466,9 +1466,9 @@ jobs:
           fi
 
       - name: Upload postpublish evidence
-        if: ${{ always() }}
+        if: ${{ always() && inputs.publish_openclaw_npm }}
         uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
         with:
           name: openclaw-release-postpublish-evidence-${{ inputs.tag }}
           path: ${{ runner.temp }}/openclaw-release-postpublish-evidence
-          if-no-files-found: ignore
+          if-no-files-found: error

.github/workflows/qa-profile-evidence.yml

@@ -89,6 +89,13 @@ jobs:
         uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3 # v9
         with:
           script: |
+            // Reusable workflow jobs inherit the caller event but run as
+            // github-actions[bot]; selected ref validation still gates secrets.
+            if (context.actor === "github-actions[bot]") {
+              core.info("Skipping manual actor permission check for a reusable workflow call.");
+              core.setOutput("authorized", "true");
+              return;
+            }
             if (context.eventName !== "workflow_dispatch") {
               core.info(`Skipping manual actor permission check for ${context.eventName}.`);
               core.setOutput("authorized", "true");
@@ -243,6 +250,9 @@ jobs:
           NODE_OPTIONS: --max-old-space-size=8192
         run: node scripts/build-all.mjs qaRuntime
 
+      - name: Ensure Playwright Chromium
+        run: node scripts/ensure-playwright-chromium.mjs
+
       - name: Run QA profile
         id: run_profile
         env:

CHANGELOG.md

@@ -37,6 +37,7 @@ This audited record covers the complete v2026.6.8..HEAD history: 423 merged PRs.
 
 #### Pull requests
 
+- **PR #92154** fix(qqbot): gate private group commands and close strict command visibility gaps. Thanks @sliverp.
 - **PR #90463** refactor: add session accessor seam with gateway consumer. Thanks @jalehman.
 - **PR #88656** Drop reasoning-only length turns from replay. Thanks @abel-zer0.
 - **PR #92856** feat(webui): add session workspace rail. Thanks @Solvely-Colin.

apps/ios/Sources/Design/SettingsProTabActions.swift

@@ -152,6 +152,7 @@ extension SettingsProTab {
         }
         let notificationSettings = await UNUserNotificationCenter.current().notificationSettings()
         self.applyNotificationStatus(notificationSettings.authorizationStatus)
+        self.registerForRemoteNotificationsIfEnrollmentReady()
 
         let issueCount = SettingsDiagnostics.issueCount(
             gatewayConnected: self.gatewayDiagnosticConnected,
@@ -417,6 +418,7 @@ extension SettingsProTab {
             let status = settings.authorizationStatus
             Task { @MainActor in
                 self.applyNotificationStatus(status)
+                self.registerForRemoteNotificationsIfEnrollmentReady()
             }
         }
     }
@@ -437,6 +439,7 @@ extension SettingsProTab {
 
     func requestNotificationAuthorizationFromSettings() {
         guard !self.isRequestingNotificationAuthorization else { return }
+        PushEnrollmentConsent.markDisclosureAccepted()
         self.isRequestingNotificationAuthorization = true
         Task {
             let granted = await (try? UNUserNotificationCenter.current().requestAuthorization(options: [
@@ -448,12 +451,19 @@ extension SettingsProTab {
             await MainActor.run {
                 self.isRequestingNotificationAuthorization = false
                 self.notificationStatus = SettingsNotificationStatus(settings.authorizationStatus)
-                guard granted, self.notificationStatus.allowsNotifications else { return }
-                UIApplication.shared.registerForRemoteNotifications()
+                guard granted else { return }
+                self.registerForRemoteNotificationsIfEnrollmentReady()
             }
         }
     }
 
+    @MainActor
+    func registerForRemoteNotificationsIfEnrollmentReady() {
+        guard PushEnrollmentConsent.disclosureAccepted else { return }
+        guard self.notificationStatus.allowsNotifications else { return }
+        UIApplication.shared.registerForRemoteNotifications()
+    }
+
     @MainActor
     func applyNotificationStatus(_ status: UNAuthorizationStatus) {
         self.notificationStatus = SettingsNotificationStatus(status)

apps/ios/Sources/Model/NodeAppModel.swift

@@ -4103,6 +4103,9 @@ extension NodeAppModel {
 
     private func registerAPNsTokenIfNeeded() async {
         let usesRelayTransport = await self.pushRegistrationManager.usesRelayTransport
+        guard await self.canPublishAPNsRegistration(usesRelayTransport: usesRelayTransport) else {
+            return
+        }
         guard self.gatewayConnected else {
             if usesRelayTransport {
                 GatewayDiagnostics.pushRelay.skipped("gateway_offline")
@@ -4163,6 +4166,23 @@ extension NodeAppModel {
         }
     }
 
+    private func canPublishAPNsRegistration(usesRelayTransport: Bool) async -> Bool {
+        guard PushEnrollmentConsent.disclosureAccepted else {
+            if usesRelayTransport {
+                GatewayDiagnostics.pushRelay.skipped("enrollment_disclosure_not_accepted")
+            }
+            return false
+        }
+        let status = await self.notificationAuthorizationStatus()
+        guard Self.isNotificationAuthorizationAllowed(status) else {
+            if usesRelayTransport {
+                GatewayDiagnostics.pushRelay.skipped("notifications_not_authorized")
+            }
+            return false
+        }
+        return true
+    }
+
     private func fetchPushRelayGatewayIdentity() async throws -> PushRelayGatewayIdentity {
         let response = try await self.operatorGateway.request(
             method: "gateway.identity.get",
@@ -5126,6 +5146,10 @@ extension NodeAppModel {
         self.setOperatorConnected(connected)
     }
 
+    func _test_canPublishAPNsRegistration(usesRelayTransport: Bool = true) async -> Bool {
+        await self.canPublishAPNsRegistration(usesRelayTransport: usesRelayTransport)
+    }
+
     nonisolated static func _test_makeWatchChatItems(from raw: [OpenClawKit.AnyCodable]) -> [OpenClawWatchChatItem] {
         self.makeWatchChatItems(from: raw)
     }

apps/ios/Sources/OpenClawApp.swift

@@ -123,10 +123,30 @@ final class OpenClawAppDelegate: NSObject, UIApplicationDelegate, @preconcurrenc
         let notificationCenter = UNUserNotificationCenter.current()
         notificationCenter.delegate = self
         ExecApprovalNotificationBridge.registerCategory(center: notificationCenter)
-        application.registerForRemoteNotifications()
+        Task { @MainActor in
+            await self.registerForRemoteNotificationsIfEnrollmentReady(application)
+        }
         return true
     }
 
+    private func registerForRemoteNotificationsIfEnrollmentReady(_ application: UIApplication) async {
+        guard PushEnrollmentConsent.disclosureAccepted else { return }
+        guard await Self.isNotificationAuthorizationAllowed() else { return }
+        application.registerForRemoteNotifications()
+    }
+
+    private static func isNotificationAuthorizationAllowed() async -> Bool {
+        let settings = await UNUserNotificationCenter.current().notificationSettings()
+        switch settings.authorizationStatus {
+        case .authorized, .provisional, .ephemeral:
+            return true
+        case .denied, .notDetermined:
+            return false
+        @unknown default:
+            return false
+        }
+    }
+
     func application(_ application: UIApplication, didRegisterForRemoteNotificationsWithDeviceToken deviceToken: Data) {
         if let appModel = self.resolvedAppModel() {
             Task { @MainActor in

apps/ios/Sources/Push/PushEnrollmentConsent.swift

@@ -0,0 +1,19 @@
+import Foundation
+
+enum PushEnrollmentConsent {
+    static let disclosureAcceptedKey = "push.enrollment.disclosureAccepted"
+
+    static var disclosureAccepted: Bool {
+        UserDefaults.standard.bool(forKey: disclosureAcceptedKey)
+    }
+
+    static func markDisclosureAccepted() {
+        UserDefaults.standard.set(true, forKey: self.disclosureAcceptedKey)
+    }
+
+    #if DEBUG
+    static func reset() {
+        UserDefaults.standard.removeObject(forKey: self.disclosureAcceptedKey)
+    }
+    #endif
+}

apps/ios/SwiftSources.input.xcfilelist

@@ -76,6 +76,7 @@ Sources/Permissions/PermissionRequestBridge.swift
 Sources/Push/ExecApprovalNotificationBridge.swift
 Sources/Push/BackgroundAliveBeacon.swift
 Sources/Push/PushBuildConfig.swift
+Sources/Push/PushEnrollmentConsent.swift
 Sources/Push/PushRegistrationManager.swift
 Sources/Push/PushRelayClient.swift
 Sources/Push/PushRelayKeychainStore.swift

apps/ios/Tests/NodeAppModelInvokeTests.swift

@@ -1377,6 +1377,24 @@ private final class MockBootstrapNotificationCenter: NotificationCentering, @unc
         #expect(center.addCalls == 1)
     }
 
+    @Test @MainActor func apnsRegistrationRequiresDisclosureAndNotificationAuthorization() async {
+        let center = MockBootstrapNotificationCenter()
+        center.status = .authorized
+        let appModel = NodeAppModel(notificationCenter: center)
+        PushEnrollmentConsent.reset()
+        defer { PushEnrollmentConsent.reset() }
+
+        #expect(await appModel._test_canPublishAPNsRegistration() == false)
+        #expect(await appModel._test_canPublishAPNsRegistration(usesRelayTransport: false) == false)
+
+        PushEnrollmentConsent.markDisclosureAccepted()
+        center.status = .notDetermined
+        #expect(await appModel._test_canPublishAPNsRegistration() == false)
+
+        center.status = .authorized
+        #expect(await appModel._test_canPublishAPNsRegistration())
+    }
+
     @Test @MainActor func chatPushWithoutSpeechReturnsUnavailableWhenNotificationsOff() async throws {
         let center = MockBootstrapNotificationCenter()
         center.status = .notDetermined

apps/ios/Tests/RootTabsSourceGuardTests.swift

@@ -550,6 +550,20 @@ struct RootTabsSourceGuardTests {
         #expect(docsSource.contains(".accessibilityHint(\"Opens Settings / Gateway\")"))
     }
 
+    @Test func `push enrollment stays behind notification disclosure flow`() throws {
+        let appSource = try String(contentsOf: Self.openClawAppSourceURL(), encoding: .utf8)
+        let actionsSource = try String(contentsOf: Self.settingsProTabActionsSourceURL(), encoding: .utf8)
+        let modelSource = try String(contentsOf: Self.nodeAppModelSourceURL(), encoding: .utf8)
+
+        #expect(appSource.contains("PushEnrollmentConsent.disclosureAccepted"))
+        #expect(appSource.contains("await Self.isNotificationAuthorizationAllowed()"))
+        #expect(actionsSource.contains("PushEnrollmentConsent.markDisclosureAccepted()"))
+        #expect(actionsSource.contains("self.registerForRemoteNotificationsIfEnrollmentReady()"))
+        #expect(modelSource.contains("PushEnrollmentConsent.disclosureAccepted"))
+        #expect(modelSource.contains("notifications_not_authorized"))
+        #expect(modelSource.contains("enrollment_disclosure_not_accepted"))
+    }
+
     @Test func `gateway settings keeps pairing trust diagnostics and tailscale actions`() throws {
         let settingsSource = try String(contentsOf: Self.settingsProTabSourceURL(), encoding: .utf8)
         let sectionsSource = try String(contentsOf: Self.settingsProTabSectionsSourceURL(), encoding: .utf8)
@@ -786,6 +800,13 @@ struct RootTabsSourceGuardTests {
             .appendingPathComponent("Sources/Design/SettingsProTab.swift")
     }
 
+    private static func openClawAppSourceURL() -> URL {
+        URL(fileURLWithPath: #filePath)
+            .deletingLastPathComponent()
+            .deletingLastPathComponent()
+            .appendingPathComponent("Sources/OpenClawApp.swift")
+    }
+
     private static func notificationPermissionGuidanceDialogSourceURL() -> URL {
         URL(fileURLWithPath: #filePath)
             .deletingLastPathComponent()

apps/ios/fastlane/Fastfile

@@ -284,6 +284,7 @@ def normalize_watch_screenshot_status_bar(path)
   script = <<~SWIFT
     import AppKit
     import Foundation
+    import ImageIO
 
     let path = CommandLine.arguments[1]
     let timeText = CommandLine.arguments[2]
@@ -295,36 +296,37 @@ def normalize_watch_screenshot_status_bar(path)
       exit(2)
     }
 
-    let width = CGFloat(cgImage.width)
-    let height = CGFloat(cgImage.height)
-    guard let bitmap = NSBitmapImageRep(
-      bitmapDataPlanes: nil,
-      pixelsWide: Int(width),
-      pixelsHigh: Int(height),
-      bitsPerSample: 8,
-      samplesPerPixel: 4,
-      hasAlpha: true,
-      isPlanar: false,
-      colorSpaceName: .deviceRGB,
-      bytesPerRow: 0,
-      bitsPerPixel: 0),
-      let graphicsContext = NSGraphicsContext(bitmapImageRep: bitmap)
+    let width = cgImage.width
+    let height = cgImage.height
+    let drawWidth = CGFloat(width)
+    let drawHeight = CGFloat(height)
+    let colorSpace = CGColorSpace(name: CGColorSpace.sRGB) ?? CGColorSpaceCreateDeviceRGB()
+    guard let bitmapContext = CGContext(
+      data: nil,
+      width: width,
+      height: height,
+      bitsPerComponent: 8,
+      bytesPerRow: width * 4,
+      space: colorSpace,
+      bitmapInfo: CGImageAlphaInfo.noneSkipLast.rawValue)
     else {
       fputs("Failed to create normalized screenshot bitmap at \\(path)\\n", stderr)
       exit(3)
     }
 
-    bitmap.size = NSSize(width: width, height: height)
+    let graphicsContext = NSGraphicsContext(cgContext: bitmapContext, flipped: false)
     NSGraphicsContext.saveGraphicsState()
     NSGraphicsContext.current = graphicsContext
+    NSColor.black.setFill()
+    NSBezierPath(rect: NSRect(x: 0, y: 0, width: drawWidth, height: drawHeight)).fill()
     source.draw(
-      in: NSRect(x: 0, y: 0, width: width, height: height),
-      from: NSRect(x: 0, y: 0, width: width, height: height),
-      operation: .copy,
+      in: NSRect(x: 0, y: 0, width: drawWidth, height: drawHeight),
+      from: NSRect(x: 0, y: 0, width: drawWidth, height: drawHeight),
+      operation: .sourceOver,
       fraction: 1.0)
 
     NSColor.black.setFill()
-    NSBezierPath(rect: NSRect(x: width - 146, y: height - 92, width: 124, height: 70)).fill()
+    NSBezierPath(rect: NSRect(x: drawWidth - 146, y: drawHeight - 92, width: 124, height: 70)).fill()
 
     let paragraphStyle = NSMutableParagraphStyle()
     paragraphStyle.alignment = .right
@@ -334,17 +336,26 @@ def normalize_watch_screenshot_status_bar(path)
       .paragraphStyle: paragraphStyle,
     ]
     timeText.draw(
-      in: NSRect(x: width - 134, y: height - 82, width: 102, height: 44),
+      in: NSRect(x: drawWidth - 134, y: drawHeight - 82, width: 102, height: 44),
       withAttributes: attributes)
     NSGraphicsContext.restoreGraphicsState()
 
-    guard let png = bitmap.representation(using: .png, properties: [:])
+    guard let output = bitmapContext.makeImage(),
+          let destination = CGImageDestinationCreateWithURL(
+            URL(fileURLWithPath: path) as CFURL,
+            "public.png" as CFString,
+            1,
+            nil)
     else {
       fputs("Failed to encode normalized screenshot at \\(path)\\n", stderr)
       exit(4)
     }
 
-    try png.write(to: URL(fileURLWithPath: path))
+    CGImageDestinationAddImage(destination, output, nil)
+    guard CGImageDestinationFinalize(destination) else {
+      fputs("Failed to write normalized screenshot at \\(path)\\n", stderr)
+      exit(5)
+    }
   SWIFT
 
   Tempfile.create(["openclaw-watch-status-bar", ".swift"]) do |file|

docs/.generated/plugin-sdk-api-baseline.sha256

@@ -1,2 +1,2 @@
-da3373338b7f9c5f5639ad8233a32897d2346a0babe69a77386a7bff154cdcb1  plugin-sdk-api-baseline.json
-17404d885e0d64ebc8e3c99443921058a8f1aebf76a5e612eb1f0cd7817d48f0  plugin-sdk-api-baseline.jsonl
+57b3f65c9d8c4edddea6ffa86584756234e761cc1cdd561e4f57c8c072baaad2  plugin-sdk-api-baseline.json
+1c20edb5599d0050382a32272ff3708e969f4605a2dca3db8b5cef9ab7680bd6  plugin-sdk-api-baseline.jsonl

docs/channels/mattermost.md

@@ -155,6 +155,7 @@ Notes:
 
 - `onchar` still responds to explicit @mentions.
 - `channels.mattermost.requireMention` is honored for legacy configs but `chatmode` is preferred.
+- After the bot sends a visible reply in a channel thread, later messages in that same thread are answered without a new @mention or `onchar` prefix, so multi-turn thread conversations keep flowing. Participation is remembered for 7 days of thread inactivity (refreshed on each reply) and persists across gateway restarts. Threads the bot has only observed are unaffected; start a new top-level message to require an explicit mention again.
 
 ## Threading and sessions
 

docs/channels/qqbot.md

@@ -151,6 +151,7 @@ to a group, then mention it or configure the group to run without a mention.
       groups: {
         "*": {
           requireMention: true,
+          commandLevel: "all",
           historyLimit: 50,
           tools: { deny: ["exec", "read", "write"] },
         },
@@ -158,6 +159,7 @@ to a group, then mention it or configure the group to run without a mention.
           name: "Release room",
           requireMention: false,
           ignoreOtherMentions: true,
+          commandLevel: "safety",
           historyLimit: 20,
           prompt: "Keep replies short and operational.",
         },
@@ -172,6 +174,9 @@ to a group, then mention it or configure the group to run without a mention.
 settings include:
 
 - `requireMention`: require an @mention before the bot replies. Default: `true`.
+- `commandLevel`: control which built-in slash commands can run in groups.
+  Default: `all`, which preserves the pre-existing QQBot group behavior when the
+  setting is omitted.
 - `ignoreOtherMentions`: drop messages that mention someone else but not the bot.
 - `historyLimit`: keep recent non-mention group messages as context for the next mentioned turn. Set `0` to disable.
 - `tools`: allow/deny tools for the whole group.
@@ -179,6 +184,17 @@ settings include:
 - `name`: friendly label used in logs and group context.
 - `prompt`: per-group behavior prompt appended to the agent context.
 
+`commandLevel` accepts:
+
+- `all`: keep recognized built-in commands available as before. Some commands may
+  stay hidden from menus, but authorized users can still run them in the group.
+- `safety`: allow common collaboration commands such as `/help`, `/btw`, and
+  `/stop`; ask users to run sensitive commands such as `/config`, `/tools`, and
+  `/bash` in private chat.
+- `strict`: only allow the group-session controls needed for strict group
+  operation. `/stop` still stays urgent so an authorized sender can interrupt an
+  active run.
+
 Old QQBot `toolPolicy` entries are retired. Run `openclaw doctor --fix` to migrate them to `tools`.
 
 Activation modes are `mention` and `always`. `requireMention: true` maps to

docs/ci.md

@@ -198,7 +198,7 @@ Every lane uploads GitHub artifacts. When `CLAWGRIT_REPORTS_TOKEN` is configured
 
 ## Full Release Validation
 
-`Full Release Validation` is the manual umbrella workflow for "run everything before release." It accepts a branch, tag, or full commit SHA, dispatches the manual `CI` workflow with that target, dispatches `Plugin Prerelease` for release-only plugin/package/static/Docker proof, and dispatches `OpenClaw Release Checks` for install smoke, package acceptance, cross-OS package checks, QA Lab parity, Matrix, and Telegram lanes. Stable and full profiles always include exhaustive live/E2E and Docker release-path soak coverage; the beta profile can opt in with `run_release_soak=true`. The canonical package Telegram E2E runs inside Package Acceptance, so a full candidate does not start a duplicate live poller. After publishing, pass `release_package_spec` to reuse the shipped npm package across release checks, Package Acceptance, Docker, cross-OS, and Telegram without rebuilding. Use `npm_telegram_package_spec` only for a focused published-package Telegram rerun. The Codex plugin live package lane uses the same selected state by default: published `release_package_spec=openclaw@<tag>` derives `codex_plugin_spec=npm:@openclaw/codex@<tag>`, while SHA/artifact runs pack `extensions/codex` from the selected ref. Set `codex_plugin_spec` explicitly for custom plugin sources such as `npm:`, `npm-pack:`, or `git:` specs.
+`Full Release Validation` is the manual umbrella workflow for "run everything before release." It accepts a branch, tag, or full commit SHA, dispatches the manual `CI` workflow with that target, dispatches `Plugin Prerelease` for release-only plugin/package/static/Docker proof, and dispatches `OpenClaw Release Checks` for install smoke, package acceptance, cross-OS package checks, maturity scorecard rendering from QA profile evidence, QA Lab parity, Matrix, and Telegram lanes. Stable and full profiles always include exhaustive live/E2E and Docker release-path soak coverage; the beta profile can opt in with `run_release_soak=true`. The canonical package Telegram E2E runs inside Package Acceptance, so a full candidate does not start a duplicate live poller. After publishing, pass `release_package_spec` to reuse the shipped npm package across release checks, Package Acceptance, Docker, cross-OS, and Telegram without rebuilding. Use `npm_telegram_package_spec` only for a focused published-package Telegram rerun. The Codex plugin live package lane uses the same selected state by default: published `release_package_spec=openclaw@<tag>` derives `codex_plugin_spec=npm:@openclaw/codex@<tag>`, while SHA/artifact runs pack `extensions/codex` from the selected ref. Set `codex_plugin_spec` explicitly for custom plugin sources such as `npm:`, `npm-pack:`, or `git:` specs.
 
 See [Full release validation](/reference/full-release-validation) for the
 stage matrix, exact workflow job names, profile differences, artifacts, and

docs/docs.json

@@ -68,6 +68,14 @@
       "source": "/reference/openclaw-sdk-api-design",
       "destination": "/gateway/external-apps"
     },
+    {
+      "source": "/reference/maturity-scorecard",
+      "destination": "/maturity/scorecard"
+    },
+    {
+      "source": "/reference/maturity-taxonomy",
+      "destination": "/maturity/taxonomy"
+    },
     {
       "source": "/mcp",
       "destination": "/cli/mcp"
@@ -1852,6 +1860,8 @@
               {
                 "group": "Release and CI",
                 "pages": [
+                  "maturity/scorecard",
+                  "maturity/taxonomy",
                   "reference/RELEASING",
                   "reference/full-release-validation",
                   "reference/release-performance-sweep",

docs/install/uninstall.md

@@ -110,14 +110,18 @@ systemctl --user daemon-reload
 ### Windows (Scheduled Task)
 
 Default task name is `OpenClaw Gateway` (or `OpenClaw Gateway (<profile>)`).
-The task script lives under your state dir.
+The task script lives under your state dir as `gateway.cmd`; current installs may
+also create a windowless `gateway.vbs` launcher that Task Scheduler runs instead
+of opening `gateway.cmd` directly.
 
 ```powershell
 schtasks /Delete /F /TN "OpenClaw Gateway"
-Remove-Item -Force "$env:USERPROFILE\.openclaw\gateway.cmd"
+Remove-Item -Force "$env:USERPROFILE\.openclaw\gateway.cmd" -ErrorAction SilentlyContinue
+Remove-Item -Force "$env:USERPROFILE\.openclaw\gateway.vbs" -ErrorAction SilentlyContinue
 ```
 
-If you used a profile, delete the matching task name and `~\.openclaw-<profile>\gateway.cmd`.
+If you used a profile, delete the matching task name and the `gateway.cmd` /
+`gateway.vbs` files under `~\.openclaw-<profile>`.
 
 ## Normal install vs source checkout
 

docs/platforms/windows.md

@@ -124,8 +124,11 @@ openclaw gateway status --json
 ```
 
 Native Windows CLI and Gateway flows are supported and continue to improve.
-Managed startup uses Windows Scheduled Tasks when available and falls back to a
-per-user Startup-folder login item if task creation is denied.
+Managed startup uses Windows Scheduled Tasks when available. The task keeps the
+readable `gateway.cmd` script in the OpenClaw state dir, but launches it through
+a generated `gateway.vbs` WScript wrapper so the background Gateway does not open
+a visible console window. If task creation is denied, OpenClaw falls back to a
+per-user Startup-folder login item.
 
 To install the Gateway service:
 

docs/plugins/copilot.md

@@ -259,14 +259,10 @@ under `describe("runSideQuestion")`.
   fallback for whatever runtimes do not have a peer surface.
 - PI session state is not migrated when an agent switches to `copilot`.
   Selection is per attempt; existing PI sessions remain valid.
-- **Interactive `ask_user` is not yet wired.** The SDK's
-  `onUserInputRequest` handler is intentionally not registered, which
-  per the SDK contract hides the `ask_user` tool from the model
-  entirely. Agents running under this harness make best-judgment
-  decisions from the initial prompt rather than asking clarifying
-  questions mid-turn. A follow-up will port the codex pattern at
-  `extensions/codex/src/app-server/user-input-bridge.ts` to route SDK
-  `UserInputRequest`s through the OpenClaw channel/TUI prompt path.
+- `ask_user` uses the same OpenClaw prompt-and-reply path as the Codex
+  harness. When the Copilot SDK asks for user input, OpenClaw posts a
+  blocking prompt to the active channel/TUI and the next queued user
+  message resolves the SDK request.
 
 ## Permissions and ask_user
 
@@ -328,11 +324,15 @@ the tool bridge. The bridge also forwards the bounded tool-construction
 controls it can enforce at the SDK boundary: `includeCoreTools`, the
 runtime tool allowlist, and `toolConstructionPlan`.
 
-The remaining PI tool-search/code-mode fields are intentionally **not**
-forwarded at MVP and tracked as follow-ups: `toolSearchCatalogRef`,
-`includeToolSearchControls`, and `toolSearchCatalogExecutor`. Those
-controls drive PI's native tool-search UI and have no direct Copilot SDK
-analog yet.
+The bridge also uses the shared harness tool-surface helper from
+`openclaw/plugin-sdk/agent-harness-tool-runtime` for PI parity. When
+tool-search is enabled, the SDK sees compact control tools plus a hidden
+catalog executor instead of every OpenClaw tool schema. When code mode is
+enabled, the helper builds the same code-mode control surface and catalog
+lifecycle used by other agent harnesses. Local-model lean defaults,
+runtime-compatible schema filtering, directory hydration, and catalog
+cleanup all stay in the shared helper so Copilot and Codex-adjacent
+harnesses do not drift.
 
 ### Session-level GitHub token
 
@@ -349,7 +349,10 @@ When the resolved mode is `useLoggedInUser`, the session-level field
 is omitted so the SDK keeps deriving identity from the logged-in
 identity.
 
-`ask_user` is intentionally hidden — see Limitations above.
+`ask_user` uses `SessionConfig.onUserInputRequest`. The bridge accepts
+choice indexes or labels for fixed-choice requests, accepts free-form
+answers when the SDK request allows them, and cancels a pending request
+when the OpenClaw attempt is aborted.
 
 ## Related
 

docs/plugins/sdk-agent-harness.md

@@ -196,6 +196,23 @@ finish. Both helpers accept the same `{ event, ctx }` payload as
 `runAgentHarnessAgentEndHook(...)`; their failures do not alter the completed
 attempt result.
 
+### User input and tool surfaces
+
+Native harnesses that expose a runtime-level user-input request should use the
+user-input helpers from `openclaw/plugin-sdk/agent-harness-runtime` to format
+the prompt, deliver it through OpenClaw's blocking reply path, and normalize
+choice/free-form answers back into the runtime's native response shape. The
+helper keeps channel/TUI presentation consistent while each harness keeps its
+own protocol parsing and pending-request lifecycle.
+
+Native harnesses that need PI-like compact tool routing should use
+`createAgentHarnessToolSurfaceRuntime(...)` from
+`openclaw/plugin-sdk/agent-harness-tool-runtime`. It owns
+tool-search/code-mode control selection, local-model lean defaults,
+runtime-compatible schema filtering, hidden catalog execution, directory
+hydration, and catalog cleanup. Harnesses still own their SDK-specific tool
+conversion and native execution callback.
+
 ### Native Codex harness mode
 
 The bundled `codex` harness is the native Codex mode for embedded OpenClaw

docs/providers/xiaomi.md

@@ -84,8 +84,8 @@ Choose the Token Plan auth choice that matches the regional base URL shown in Xi
 
 | Model ref                         | Input       | Context   | Max output | Reasoning | Notes         |
 | --------------------------------- | ----------- | --------- | ---------- | --------- | ------------- |
-| `xiaomi-token-plan/mimo-v2.5-pro` | text        | 1,048,576 | 32,000     | Yes       | Default model |
-| `xiaomi-token-plan/mimo-v2.5`     | text, image | 1,048,576 | 32,000     | Yes       | Multimodal    |
+| `xiaomi-token-plan/mimo-v2.5-pro` | text        | 1,048,576 | 131,072    | Yes       | Default model |
+| `xiaomi-token-plan/mimo-v2.5`     | text, image | 1,048,576 | 131,072    | Yes       | Multimodal    |
 
 <Tip>
 Token Plan onboarding validates the key shape and warns when a `tp-...` key is entered into the pay-as-you-go path, or an `sk-...` key is entered into the Token Plan path.
@@ -222,7 +222,7 @@ Token Plan:
             reasoning: true,
             input: ["text"],
             contextWindow: 1048576,
-            maxTokens: 32000,
+            maxTokens: 131072,
           },
           {
             id: "mimo-v2.5",
@@ -230,7 +230,7 @@ Token Plan:
             reasoning: true,
             input: ["text", "image"],
             contextWindow: 1048576,
-            maxTokens: 32000,
+            maxTokens: 131072,
           },
         ],
       },

docs/style.css

@@ -135,3 +135,753 @@ html.dark .nav-tabs-underline {
     grid-template-columns: 1fr;
   }
 }
+
+.maturity-hero {
+  display: grid;
+  gap: 14px;
+  margin: 10px 0 38px;
+  padding: 4px 0 26px 20px;
+  border-bottom: 1px solid color-mix(in oklab, rgb(var(--primary)) 22%, transparent);
+  border-left: 3px solid rgb(var(--primary));
+}
+
+.maturity-hero-compact {
+  margin-bottom: 34px;
+}
+
+.maturity-hero h2 {
+  max-width: 46rem;
+  margin: 0;
+  font-size: clamp(26px, 3vw, 38px);
+  line-height: 1.08;
+  letter-spacing: -0.01em;
+}
+
+.maturity-hero-title {
+  max-width: 46rem;
+  margin: 0;
+  font-size: clamp(26px, 3vw, 38px);
+  font-weight: 750;
+  line-height: 1.08;
+  letter-spacing: -0.01em;
+}
+
+.maturity-hero > p:not(.maturity-kicker):not(.maturity-jump-links) {
+  max-width: 58rem;
+  margin: 0;
+  font-size: 16px;
+  line-height: 1.65;
+  opacity: 0.76;
+}
+
+.maturity-kicker {
+  margin: 0;
+  color: rgb(var(--primary));
+  font-size: 10px;
+  font-weight: 750;
+  letter-spacing: 0.1em;
+  line-height: 1.3;
+  text-transform: uppercase;
+}
+
+.maturity-jump-links {
+  margin: 0;
+  font-size: 13px;
+  line-height: 1.5;
+}
+
+.maturity-jump-links a {
+  text-decoration: none;
+}
+
+.maturity-jump-links a:hover {
+  text-decoration: underline;
+  text-underline-offset: 3px;
+}
+
+.maturity-score-stable,
+.maturity-band-stable {
+  color: #4ca574;
+}
+
+.maturity-score-beta,
+.maturity-band-beta {
+  color: #849fd2;
+}
+
+.maturity-score-alpha,
+.maturity-band-alpha {
+  color: #d39a4b;
+}
+
+.maturity-score-experimental,
+.maturity-band-experimental {
+  color: #dc7669;
+}
+
+.maturity-score-clawesome,
+.maturity-band-clawesome {
+  color: #46b59a;
+}
+
+.maturity-level-pill {
+  display: inline-flex;
+  align-items: center;
+  gap: 6px;
+  width: max-content;
+  max-width: 100%;
+  padding: 3px 8px;
+  border: 1px solid color-mix(in oklab, currentColor 32%, transparent);
+  border-radius: 999px;
+  background: color-mix(in oklab, currentColor 10%, transparent);
+  color: inherit;
+  font-size: 10px;
+  font-weight: 750;
+  line-height: 1.25;
+  white-space: nowrap;
+}
+
+.maturity-level-code {
+  font-size: 9px;
+  letter-spacing: 0.04em;
+  opacity: 0.72;
+}
+
+.maturity-level-experimental {
+  color: #dc7669;
+}
+
+.maturity-level-alpha {
+  color: #d39a4b;
+}
+
+.maturity-level-beta {
+  color: #849fd2;
+}
+
+.maturity-level-stable {
+  color: #4ca574;
+}
+
+.maturity-level-clawesome {
+  color: #46b59a;
+}
+
+.maturity-summary-grid {
+  display: grid;
+  grid-template-columns: repeat(3, minmax(0, 1fr));
+  margin: 14px 0 20px;
+  border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 18%, transparent);
+  border-bottom: 1px solid color-mix(in oklab, rgb(var(--primary)) 18%, transparent);
+}
+
+.maturity-summary-item {
+  display: grid;
+  gap: 10px;
+  min-width: 0;
+  padding: 18px 20px 18px 0;
+}
+
+.maturity-summary-item + .maturity-summary-item {
+  padding-left: 20px;
+  border-left: 1px solid color-mix(in oklab, rgb(var(--primary)) 14%, transparent);
+}
+
+.maturity-summary-heading {
+  display: flex;
+  align-items: baseline;
+  gap: 9px;
+}
+
+.maturity-summary-value {
+  display: inline-block;
+  font-size: 30px;
+  font-weight: 750;
+  letter-spacing: -0.04em;
+}
+
+.maturity-summary-heading > span:not(.maturity-summary-value) {
+  font-size: 13px;
+  font-weight: 700;
+}
+
+.maturity-summary-bar {
+  height: 7px;
+  overflow: hidden;
+  background: color-mix(in oklab, currentColor 14%, transparent);
+}
+
+.maturity-summary-bar span {
+  display: block;
+  width: calc(var(--score) * 1%);
+  height: 100%;
+  background: currentColor;
+}
+
+.maturity-summary-meta {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 4px 10px;
+  font-size: 11px;
+  line-height: 1.4;
+}
+
+.maturity-summary-meta span:first-child {
+  font-weight: 700;
+}
+
+.maturity-summary-meta span:last-child {
+  opacity: 0.62;
+}
+
+.maturity-band-list {
+  display: flex;
+  margin: 12px 0 30px;
+  border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 16%, transparent);
+  border-bottom: 1px solid color-mix(in oklab, rgb(var(--primary)) 16%, transparent);
+}
+
+.maturity-band {
+  display: grid;
+  flex: 1 1 0;
+  gap: 3px;
+  padding: 10px 12px 11px 0;
+}
+
+.maturity-band + .maturity-band {
+  padding-left: 12px;
+  border-left: 1px solid color-mix(in oklab, rgb(var(--primary)) 12%, transparent);
+}
+
+.maturity-band-title {
+  font-size: 12px;
+  font-weight: 700;
+}
+
+.maturity-band-title + span {
+  color: inherit;
+  font-size: 11px;
+  opacity: 0.7;
+}
+
+.maturity-band > span:last-child {
+  color: inherit;
+  font-size: 11px;
+  opacity: 0.7;
+}
+
+.maturity-band span {
+  color: inherit;
+  font-size: 11px;
+  opacity: 0.7;
+}
+
+.maturity-band .maturity-level-pill {
+  font-size: 10px;
+  opacity: 1;
+}
+
+.maturity-band .maturity-level-pill span {
+  font-size: inherit;
+  opacity: inherit;
+}
+
+.maturity-score {
+  display: grid;
+  gap: 4px;
+  min-width: 0;
+  color: inherit;
+  font-size: 11px;
+  font-weight: 700;
+}
+
+.maturity-score-label {
+  display: flex;
+  justify-content: space-between;
+  gap: 6px;
+  line-height: 1.2;
+}
+
+.maturity-score-label > span:first-child {
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.maturity-score-label > span:last-child {
+  flex: 0 0 auto;
+}
+
+.maturity-score-label .maturity-level-pill {
+  gap: 4px;
+  padding: 2px 6px;
+  font-size: 9px;
+}
+
+.maturity-score-label-text {
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.maturity-summary-meta .maturity-level-pill {
+  opacity: 1;
+}
+
+.maturity-meter {
+  display: inline-block;
+  width: 100%;
+  height: 3px;
+  overflow: hidden;
+  background: color-mix(in oklab, currentColor 15%, transparent);
+  vertical-align: middle;
+}
+
+.maturity-meter > span {
+  display: block;
+  width: 0;
+  height: 100%;
+  background: currentColor;
+}
+
+.maturity-score-unscored,
+.maturity-lts-none {
+  color: inherit;
+  opacity: 0.52;
+}
+
+.maturity-surface-table {
+  display: grid;
+  gap: 0;
+  margin: 8px 0 22px;
+}
+
+.maturity-surface-row {
+  display: grid;
+  grid-template-columns: minmax(190px, 1.55fr) repeat(3, minmax(110px, 1fr)) minmax(72px, 0.55fr);
+  gap: 12px;
+  align-items: center;
+  padding: 13px 0;
+  border-top: 1px solid color-mix(in oklab, currentColor 14%, transparent);
+}
+
+.maturity-surface-row-header {
+  padding: 0 0 9px;
+  border-top: 0;
+  color: inherit;
+  font-size: 10px;
+  font-weight: 750;
+  letter-spacing: 0.04em;
+  opacity: 0.56;
+  text-transform: uppercase;
+}
+
+.maturity-surface-name {
+  display: grid;
+  gap: 3px;
+  min-width: 0;
+  border-bottom: 0 !important;
+  text-decoration: none;
+}
+
+.maturity-surface-name:hover .maturity-surface-title {
+  color: rgb(var(--primary));
+}
+
+.maturity-surface-title {
+  overflow-wrap: anywhere;
+  font-size: 13px;
+  font-weight: 700;
+  line-height: 1.25;
+}
+
+.maturity-surface-meta {
+  display: flex;
+  align-items: center;
+  flex-wrap: wrap;
+  gap: 6px;
+  min-width: 0;
+}
+
+.maturity-surface-meta > span:not(.maturity-level-pill) {
+  font-size: 11px;
+  opacity: 0.62;
+}
+
+.maturity-surface-meta .maturity-level-pill {
+  opacity: 1;
+}
+
+.maturity-surface-metric {
+  min-width: 0;
+}
+
+.maturity-surface-metric-label {
+  display: none;
+  font-size: 10px;
+  opacity: 0.62;
+}
+
+.maturity-surface-support {
+  justify-self: start;
+}
+
+.maturity-lts {
+  display: inline-flex;
+  align-items: center;
+  gap: 5px;
+  color: inherit;
+  font-size: 11px;
+  font-weight: 700;
+  white-space: nowrap;
+}
+
+.maturity-lts::before {
+  width: 6px;
+  height: 6px;
+  border-radius: 50%;
+  background: currentColor;
+  content: "";
+}
+
+.maturity-lts-partial {
+  color: #d39a4b;
+}
+
+.maturity-lts-full {
+  color: #4ca574;
+}
+
+.maturity-evidence-grid {
+  display: grid;
+  grid-template-columns: repeat(3, minmax(0, 1fr));
+  margin: 14px 0 24px;
+  border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 16%, transparent);
+  border-bottom: 1px solid color-mix(in oklab, rgb(var(--primary)) 16%, transparent);
+}
+
+.maturity-evidence-card {
+  display: grid;
+  gap: 4px;
+  min-width: 0;
+  padding: 14px 16px 14px 0;
+}
+
+.maturity-evidence-card + .maturity-evidence-card {
+  padding-left: 16px;
+  border-left: 1px solid color-mix(in oklab, rgb(var(--primary)) 12%, transparent);
+}
+
+.maturity-evidence-title {
+  font-size: 13px;
+  font-weight: 700;
+}
+
+.maturity-evidence-card span {
+  font-size: 11px;
+  line-height: 1.4;
+  opacity: 0.68;
+}
+
+.maturity-readiness-summary {
+  margin: 0 0 12px;
+  font-size: 12px;
+  opacity: 0.65;
+}
+
+.maturity-readiness-list {
+  display: grid;
+  margin: 0;
+  border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 14%, transparent);
+}
+
+.maturity-readiness-row {
+  display: grid;
+  grid-template-columns: minmax(0, 1.7fr) minmax(120px, 0.8fr) minmax(110px, 0.7fr);
+  gap: 14px;
+  align-items: center;
+  padding: 11px 0;
+  border-bottom: 1px solid color-mix(in oklab, rgb(var(--primary)) 10%, transparent);
+  font-size: 11px;
+}
+
+.maturity-readiness-row-header {
+  padding: 8px 0;
+  border-bottom-color: color-mix(in oklab, rgb(var(--primary)) 14%, transparent);
+  font-size: 10px;
+  font-weight: 750;
+  letter-spacing: 0.04em;
+  opacity: 0.56;
+  text-transform: uppercase;
+}
+
+.maturity-readiness-area {
+  display: grid;
+  gap: 3px;
+  min-width: 0;
+}
+
+.maturity-readiness-title {
+  overflow-wrap: anywhere;
+  font-weight: 700;
+}
+
+.maturity-readiness-status {
+  font-size: 10px;
+  opacity: 0.62;
+}
+
+.maturity-readiness-status-ready {
+  color: #4ca574;
+}
+
+.maturity-readiness-status-partially-reviewed {
+  color: #d39a4b;
+}
+
+.maturity-readiness-status-needs-review {
+  color: #dc7669;
+}
+
+.maturity-category-list {
+  display: grid;
+  width: 100%;
+  margin-top: 4px;
+  overflow: hidden;
+}
+
+.maturity-category-row {
+  display: grid;
+  grid-template-columns: minmax(180px, 1.55fr) repeat(3, minmax(100px, 1fr)) minmax(140px, 1.2fr);
+  gap: 12px;
+  align-items: center;
+  padding: 12px 0;
+  border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 11%, transparent);
+  font-size: 11px;
+}
+
+.maturity-category-row-header {
+  padding: 8px 0;
+  border-top: 0;
+  color: inherit;
+  font-size: 10px;
+  font-weight: 750;
+  letter-spacing: 0.04em;
+  opacity: 0.56;
+  text-transform: uppercase;
+}
+
+.maturity-category-area {
+  display: grid;
+  gap: 3px;
+  min-width: 0;
+}
+
+.maturity-category-title {
+  overflow-wrap: anywhere;
+  font-weight: 700;
+}
+
+.maturity-category-area > span:last-child {
+  font-size: 10px;
+  opacity: 0.62;
+}
+
+.maturity-category-docs {
+  min-width: 0;
+  overflow-wrap: anywhere;
+  line-height: 1.4;
+}
+
+.maturity-category-docs a {
+  text-decoration: none;
+}
+
+.maturity-category-docs a:hover {
+  text-decoration: underline;
+  text-underline-offset: 3px;
+}
+
+.maturity-level-list {
+  display: grid;
+  margin: 12px 0 28px;
+  border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 16%, transparent);
+  border-bottom: 1px solid color-mix(in oklab, rgb(var(--primary)) 16%, transparent);
+}
+
+.maturity-level-row {
+  display: grid;
+  grid-template-columns: minmax(130px, 0.32fr) minmax(0, 1fr);
+  gap: 4px 14px;
+  padding: 13px 0;
+  border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 11%, transparent);
+}
+
+.maturity-level-row:first-child {
+  border-top: 0;
+}
+
+.maturity-level-title {
+  grid-row: span 2;
+  font-size: 13px;
+  font-weight: 700;
+}
+
+.maturity-level-title .maturity-level-pill {
+  opacity: 1;
+}
+
+.maturity-level-row span,
+.maturity-level-promotion {
+  font-size: 12px;
+  line-height: 1.45;
+  opacity: 0.68;
+}
+
+.maturity-surface-link {
+  display: grid;
+  gap: 3px;
+  margin: 0;
+  padding: 11px 0;
+  border-bottom: 1px solid color-mix(in oklab, currentColor 14%, transparent);
+  text-decoration: none;
+}
+
+.maturity-surface-link:hover {
+  color: rgb(var(--primary));
+}
+
+.maturity-surface-link .maturity-surface-title {
+  font-size: 13px;
+  font-weight: 700;
+}
+
+.maturity-surface-link > .maturity-surface-meta {
+  display: flex;
+  align-items: center;
+  flex-wrap: wrap;
+  gap: 6px;
+}
+
+.maturity-surface-link > .maturity-surface-meta > span:not(.maturity-level-pill) {
+  font-size: 11px;
+  opacity: 0.68;
+}
+
+.maturity-surface-link > .maturity-surface-meta .maturity-level-pill {
+  opacity: 1;
+}
+
+.maturity-surface-rollup {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 5px 14px;
+  margin: 0 0 14px;
+  padding: 9px 0;
+  border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 13%, transparent);
+  border-bottom: 1px solid color-mix(in oklab, rgb(var(--primary)) 13%, transparent);
+}
+
+.maturity-surface-rollup > span {
+  color: inherit;
+  font-size: 11px;
+  font-weight: 700;
+}
+
+#content table .maturity-score,
+#content table .maturity-lts {
+  white-space: nowrap;
+}
+
+@media (max-width: 960px) {
+  .maturity-summary-grid,
+  .maturity-evidence-grid {
+    grid-template-columns: 1fr;
+  }
+
+  .maturity-summary-item,
+  .maturity-summary-item + .maturity-summary-item,
+  .maturity-evidence-card,
+  .maturity-evidence-card + .maturity-evidence-card {
+    padding: 14px 0;
+    border-left: 0;
+  }
+
+  .maturity-summary-item + .maturity-summary-item,
+  .maturity-evidence-card + .maturity-evidence-card {
+    border-top: 1px solid color-mix(in oklab, rgb(var(--primary)) 12%, transparent);
+  }
+
+  .maturity-surface-row {
+    grid-template-columns: minmax(160px, 1.35fr) repeat(3, minmax(98px, 1fr)) minmax(70px, 0.5fr);
+    gap: 8px;
+  }
+
+  .maturity-category-row {
+    grid-template-columns: minmax(160px, 1.35fr) repeat(3, minmax(86px, 1fr)) minmax(110px, 1fr);
+    gap: 8px;
+  }
+}
+
+@media (max-width: 640px) {
+  .maturity-hero {
+    padding-left: 14px;
+  }
+
+  .maturity-surface-row-header {
+    display: none;
+  }
+
+  .maturity-surface-row {
+    grid-template-columns: minmax(0, 1fr) minmax(0, 1fr);
+    gap: 9px 12px;
+  }
+
+  .maturity-surface-metric {
+    display: grid;
+    grid-template-columns: auto 1fr;
+    align-items: center;
+    gap: 6px;
+  }
+
+  .maturity-surface-metric-label {
+    display: block;
+  }
+
+  .maturity-surface-support {
+    justify-self: end;
+  }
+
+  .maturity-readiness-row,
+  .maturity-category-row {
+    grid-template-columns: 1fr;
+    gap: 5px;
+  }
+
+  .maturity-readiness-row-header,
+  .maturity-category-row-header {
+    display: none;
+  }
+
+  .maturity-category-docs {
+    padding-top: 3px;
+  }
+
+  .maturity-band-list {
+    display: grid;
+    grid-template-columns: repeat(2, minmax(0, 1fr));
+  }
+
+  .maturity-band + .maturity-band {
+    padding-left: 0;
+    border-left: 0;
+  }
+
+  .maturity-level-row {
+    grid-template-columns: 1fr;
+  }
+
+  .maturity-level-title {
+    grid-row: auto;
+  }
+}

extensions/acpx/npm-shrinkwrap.json

@@ -10,7 +10,7 @@
       "dependencies": {
         "@agentclientprotocol/claude-agent-acp": "0.39.0",
         "@zed-industries/codex-acp": "0.15.0",
-        "acpx": "0.10.0",
+        "acpx": "0.11.2",
         "zod": "4.4.3"
       }
     },
@@ -196,9 +196,9 @@
       }
     },
     "node_modules/@clack/core": {
-      "version": "1.4.1",
-      "resolved": "https://registry.npmjs.org/@clack/core/-/core-1.4.1.tgz",
-      "integrity": "sha512-FILJa1gGKEFTGZAJE9RpVhrjKz3c3h4ar60dSv6cGuDqufQ84YEIS3GAGvZiN+H6yaLbbvTFNejjCC4tXpZEuw==",
+      "version": "1.3.1",
+      "resolved": "https://registry.npmjs.org/@clack/core/-/core-1.3.1.tgz",
+      "integrity": "sha512-fT1qHVGAag4IEkrupZ6lRRbNCs1vS9P01KB/sG8zKgvUztbYtFBtQpjSITNwooDZ83tpsPzP0mRNs1/KVszCRA==",
       "license": "MIT",
       "dependencies": {
         "fast-wrap-ansi": "^0.2.0",
@@ -209,12 +209,12 @@
       }
     },
     "node_modules/@clack/prompts": {
-      "version": "1.5.1",
-      "resolved": "https://registry.npmjs.org/@clack/prompts/-/prompts-1.5.1.tgz",
-      "integrity": "sha512-zccHj2z2oCCO4yrDiRSlFOxWerGqRiysP7a5jPK6uoI9URKAquwY42Dd/iUP8JWHxEzdRe4TlbvZCo8z1/mhrw==",
+      "version": "1.4.0",
+      "resolved": "https://registry.npmjs.org/@clack/prompts/-/prompts-1.4.0.tgz",
+      "integrity": "sha512-S0My7XPGIgpRWMDG8uRqalbgT+a6FmCUdOW+HaIOVVpUPHOb7RrpvjTjiODadKp06fsrVDJZlIzc6yCTp4AnxA==",
       "license": "MIT",
       "dependencies": {
-        "@clack/core": "1.4.1",
+        "@clack/core": "1.3.1",
         "fast-string-width": "^3.0.2",
         "fast-wrap-ansi": "^0.2.0",
         "sisteransi": "^1.0.5"
@@ -701,6 +701,7 @@
       "version": "0.15.0",
       "resolved": "https://registry.npmjs.org/@zed-industries/codex-acp/-/codex-acp-0.15.0.tgz",
       "integrity": "sha512-eAv7sGBeiYrYkOulF729nrM51szS7WIhBtugRj5wWq6csRKZUhAZfoUZlF8xUWdHPtOIzd/eT6MNG6gMHu6z0w==",
+      "deprecated": "This package has been replaced by @agentclientprotocol/codex-acp. Please migrate to continue receiving updates.",
       "license": "Apache-2.0",
       "bin": {
         "codex-acp": "bin/codex-acp.js"
@@ -721,6 +722,7 @@
       "cpu": [
         "arm64"
       ],
+      "deprecated": "This package has been replaced by @agentclientprotocol/codex-acp. Please migrate to continue receiving updates.",
       "license": "Apache-2.0",
       "optional": true,
       "os": [
@@ -737,6 +739,7 @@
       "cpu": [
         "x64"
       ],
+      "deprecated": "This package has been replaced by @agentclientprotocol/codex-acp. Please migrate to continue receiving updates.",
       "license": "Apache-2.0",
       "optional": true,
       "os": [
@@ -753,6 +756,7 @@
       "cpu": [
         "arm64"
       ],
+      "deprecated": "This package has been replaced by @agentclientprotocol/codex-acp. Please migrate to continue receiving updates.",
       "license": "Apache-2.0",
       "optional": true,
       "os": [
@@ -769,6 +773,7 @@
       "cpu": [
         "x64"
       ],
+      "deprecated": "This package has been replaced by @agentclientprotocol/codex-acp. Please migrate to continue receiving updates.",
       "license": "Apache-2.0",
       "optional": true,
       "os": [
@@ -785,6 +790,7 @@
       "cpu": [
         "arm64"
       ],
+      "deprecated": "This package has been replaced by @agentclientprotocol/codex-acp. Please migrate to continue receiving updates.",
       "license": "Apache-2.0",
       "optional": true,
       "os": [
@@ -801,6 +807,7 @@
       "cpu": [
         "x64"
       ],
+      "deprecated": "This package has been replaced by @agentclientprotocol/codex-acp. Please migrate to continue receiving updates.",
       "license": "Apache-2.0",
       "optional": true,
       "os": [
@@ -824,15 +831,15 @@
       }
     },
     "node_modules/acpx": {
-      "version": "0.10.0",
-      "resolved": "https://registry.npmjs.org/acpx/-/acpx-0.10.0.tgz",
-      "integrity": "sha512-hd48XV03gG3sd409T1lDrOKJTTz1ap4g0wrndXjxQ590tN85pBYlvfNLyerybvGRrtUGsZjNdt99r1jpIt6ukA==",
+      "version": "0.11.2",
+      "resolved": "https://registry.npmjs.org/acpx/-/acpx-0.11.2.tgz",
+      "integrity": "sha512-ksTmfJDVqUAJJXsNDamEno03AMZ/aAZzXk/h5nt61VsLc/jcpoDMfCVpErzuYNJjwCd0V6Zm5o6F8OoqxsjQWA==",
       "license": "MIT",
       "dependencies": {
-        "@agentclientprotocol/sdk": "^0.22.1",
-        "commander": "^14.0.3",
-        "skillflag": "^0.1.4",
-        "tsx": "^4.22.0",
+        "@agentclientprotocol/sdk": "^0.28.1",
+        "commander": "^15.0.0",
+        "skillflag": "^0.2.0",
+        "tsx": "^4.22.4",
         "zod": "^4.4.3"
       },
       "bin": {
@@ -842,6 +849,15 @@
         "node": ">=22.13.0"
       }
     },
+    "node_modules/acpx/node_modules/@agentclientprotocol/sdk": {
+      "version": "0.28.1",
+      "resolved": "https://registry.npmjs.org/@agentclientprotocol/sdk/-/sdk-0.28.1.tgz",
+      "integrity": "sha512-Z2Frs6YtPhnZZ+XwFXyQkRDXY0fn8FjCalEs0W4yUhQnY4TztmNq0/RnfzWdFN3vqT3h0jTz5klzYbZHGxCDyQ==",
+      "license": "Apache-2.0",
+      "peerDependencies": {
+        "zod": "^3.25.0 || ^4.0.0"
+      }
+    },
     "node_modules/ajv": {
       "version": "8.20.0",
       "resolved": "https://registry.npmjs.org/ajv/-/ajv-8.20.0.tgz",
@@ -1043,12 +1059,12 @@
       }
     },
     "node_modules/commander": {
-      "version": "14.0.3",
-      "resolved": "https://registry.npmjs.org/commander/-/commander-14.0.3.tgz",
-      "integrity": "sha512-H+y0Jo/T1RZ9qPP4Eh1pkcQcLRglraJaSLoyOtHxu6AapkjWVCy2Sit1QQ4x3Dng8qDlSsZEet7g5Pq06MvTgw==",
+      "version": "15.0.0",
+      "resolved": "https://registry.npmjs.org/commander/-/commander-15.0.0.tgz",
+      "integrity": "sha512-z67u4ZhzCL/Tydu1lJARtEZYWbWaN7oYLHbsuzocr6y4N6WZAagG3RQ4FW61V1/0+jImpj293XfrcYnd1qxtPg==",
       "license": "MIT",
       "engines": {
-        "node": ">=20"
+        "node": ">=22.12.0"
       }
     },
     "node_modules/content-disposition": {
@@ -2045,9 +2061,9 @@
       "license": "MIT"
     },
     "node_modules/skillflag": {
-      "version": "0.1.4",
-      "resolved": "https://registry.npmjs.org/skillflag/-/skillflag-0.1.4.tgz",
-      "integrity": "sha512-egFg+XCF5sloOWdtzxZivTX7n4UDj5pxQoY33wbT8h+YSDjMQJ76MZUg2rXQIBXmIDtlZhLgirS1g/3R5/qaHA==",
+      "version": "0.2.0",
+      "resolved": "https://registry.npmjs.org/skillflag/-/skillflag-0.2.0.tgz",
+      "integrity": "sha512-7ZmEpBeEoPLc+hqZ/StAnCO/hulgEPANzPyZgOM/CZ5zc3b0ApSp3URavY5POM/OKyi5d9+UC/Q21OoiYC2kJw==",
       "license": "MIT",
       "dependencies": {
         "@clack/prompts": "^1.0.1",

extensions/acpx/package.json

@@ -10,7 +10,7 @@
   "dependencies": {
     "@agentclientprotocol/claude-agent-acp": "0.39.0",
     "@zed-industries/codex-acp": "0.15.0",
-    "acpx": "0.10.0",
+    "acpx": "0.11.2",
     "zod": "4.4.3"
   },
   "devDependencies": {

extensions/acpx/src/codex-auth-bridge.test.ts

@@ -251,6 +251,15 @@ describe("prepareAcpxCodexAuthConfig", () => {
     expect(wrapper).not.toMatch(
       /forceKillTimer = setTimeout\(\(\) => killChildTree\("SIGKILL"\), 1_500\);\s*forceKillTimer\.unref\?\.\(\);\s*process\.exit\(1\);/s,
     );
+    // Orphan detection must trigger on any PPID change, not only when the new
+    // PPID is init (1). Systemd user services and container init reparent
+    // orphaned processes to a session manager or container init (PID != 1),
+    // and the older `process.ppid !== 1` guard would silently leak the codex
+    // adapter tree there.
+    expect(wrapper).not.toContain("process.ppid !== 1");
+    expect(wrapper).toMatch(
+      /setInterval\(\(\) => \{[\s\S]*?if \(process\.ppid === originalParentPid\) \{\s*return;\s*\}/,
+    );
   });
 
   it("uses the bundled Claude ACP dependency by default when it is installed", async () => {

extensions/acpx/src/codex-auth-bridge.ts

@@ -475,7 +475,13 @@ const parentWatcher =
   process.platform === "win32"
     ? undefined
     : setInterval(() => {
-        if (process.ppid === originalParentPid || process.ppid !== 1) {
+        // Orphan detection: parent PID changed means our original parent died.
+        // The new parent could be PID 1 (init) on bare-metal hosts, OR a
+        // systemd user-session manager, OR a container init, OR a session
+        // leader — depending on environment. Previously this only triggered
+        // on PPID == 1, which missed all systemd-managed deployments and
+        // leaked codex-acp adapter trees on every gateway restart.
+        if (process.ppid === originalParentPid) {
           return;
         }
         if (orphanCleanupStarted) {

extensions/acpx/src/runtime.test.ts

@@ -2,6 +2,7 @@
 import fs from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
+import { RequestedModelUnsupportedError } from "acpx/runtime";
 import { beforeEach, describe, expect, it, vi } from "vitest";
 import {
   AcpRuntimeError,
@@ -708,6 +709,100 @@ describe("AcpxRuntime fresh reset wrapper", () => {
     });
   });
 
+  it("retries without a model when ACPX reports missing model capability", async () => {
+    const baseStore: TestSessionStore = {
+      load: vi.fn(async () => undefined),
+      save: vi.fn(async () => {}),
+    };
+    const { runtime, delegate } = makeRuntime(baseStore, {
+      agentRegistry: {
+        resolve: (agentName: string) => (agentName === "opencode" ? "opencode acp" : agentName),
+        list: () => ["opencode"],
+      },
+    });
+    const ensure = vi
+      .spyOn(delegate, "ensureSession")
+      .mockRejectedValueOnce(
+        new RequestedModelUnsupportedError(
+          "Cannot apply --model: the ACP agent did not advertise model support",
+          "missing-capability",
+        ),
+      )
+      .mockResolvedValueOnce({
+        sessionKey: "agent:opencode:acp:test",
+        backend: "acpx",
+        runtimeSessionName: "opencode",
+      });
+
+    await runtime.ensureSession({
+      sessionKey: "agent:opencode:acp:test",
+      agent: "opencode",
+      mode: "persistent",
+      model: "openrouter/owl-alpha",
+    });
+
+    expect(ensure).toHaveBeenCalledTimes(2);
+    expect(readFirstEnsureSessionInput(ensure)).toMatchObject({
+      model: "openrouter/owl-alpha",
+      sessionOptions: { model: "openrouter/owl-alpha" },
+    });
+    const [, secondCall] = ensure.mock.calls;
+    expect(secondCall?.[0]).not.toHaveProperty("sessionOptions");
+    expect((secondCall?.[0] as { model?: string } | undefined)?.model).toBeUndefined();
+  });
+
+  it("does not retry when ACPX rejects an explicitly unsupported model id", async () => {
+    const baseStore: TestSessionStore = {
+      load: vi.fn(async () => undefined),
+      save: vi.fn(async () => {}),
+    };
+    const { runtime, delegate } = makeRuntime(baseStore, {
+      agentRegistry: {
+        resolve: (agentName: string) => (agentName === "opencode" ? "opencode acp" : agentName),
+        list: () => ["opencode"],
+      },
+    });
+    const ensure = vi
+      .spyOn(delegate, "ensureSession")
+      .mockRejectedValueOnce(
+        new RequestedModelUnsupportedError(
+          "Cannot apply --model: the ACP agent did not advertise that model",
+          "unadvertised-model",
+        ),
+      );
+
+    await expect(
+      runtime.ensureSession({
+        sessionKey: "agent:opencode:acp:test",
+        agent: "opencode",
+        mode: "persistent",
+        model: "unknown/model",
+      }),
+    ).rejects.toThrow("did not advertise that model");
+    expect(ensure).toHaveBeenCalledTimes(1);
+  });
+
+  it("does not retry an unrelated error with similar wording", async () => {
+    const baseStore: TestSessionStore = {
+      load: vi.fn(async () => undefined),
+      save: vi.fn(async () => {}),
+    };
+    const { runtime, delegate } = makeRuntime(baseStore);
+    const ensure = vi
+      .spyOn(delegate, "ensureSession")
+      .mockRejectedValueOnce(new Error("the ACP agent did not advertise model support"));
+
+    await expect(
+      runtime.ensureSession({
+        sessionKey: "agent:main:acp:test",
+        agent: "main",
+        mode: "persistent",
+        model: "openrouter/owl-alpha",
+      }),
+    ).rejects.toThrow("did not advertise model support");
+    expect(ensure).toHaveBeenCalledTimes(1);
+  });
+
   it("injects Codex ACP startup config into the scoped registry", () => {
     expect(testing.isCodexAcpCommand(CODEX_ACP_COMMAND)).toBe(true);
     expect(testing.isCodexAcpCommand(CODEX_ACP_WRAPPER_COMMAND)).toBe(true);

extensions/acpx/src/runtime.ts

@@ -13,6 +13,7 @@ import {
   createFileSessionStore,
   decodeAcpxRuntimeHandleState,
   encodeAcpxRuntimeHandleState,
+  isRequestedModelUnsupportedError,
   type AcpAgentRegistry,
   type AcpRuntimeDoctorReport,
   type AcpRuntimeEvent,
@@ -586,6 +587,26 @@ function withAcpxSessionOptions(input: OpenClawRuntimeEnsureInput): AcpxDelegate
   } as AcpxDelegateEnsureInput;
 }
 
+function isAcpModelCapabilityMissingError(error: unknown): boolean {
+  return isRequestedModelUnsupportedError(error) && error.reason === "missing-capability";
+}
+
+// ACPX owns the distinction between missing model capability and an invalid model id.
+// Retry only the former so explicit model mistakes remain visible to the caller.
+async function ensureDelegateSessionWithModelFallback(
+  delegate: BaseAcpxRuntime,
+  input: OpenClawRuntimeEnsureInput,
+): Promise<AcpRuntimeHandle> {
+  try {
+    return await delegate.ensureSession(withAcpxSessionOptions(input));
+  } catch (error) {
+    if (!input.model || !isAcpModelCapabilityMissingError(error)) {
+      throw error;
+    }
+    return await delegate.ensureSession(withAcpxSessionOptions({ ...input, model: undefined }));
+  }
+}
+
 function quoteShellArg(value: string): string {
   if (/^[A-Za-z0-9_./:=@+-]+$/.test(value)) {
     return value;
@@ -989,7 +1010,7 @@ export class AcpxRuntime implements AcpRuntime {
           this.withCodexWrapperDiagnostics({
             command: stableLaunchCommand,
             fallbackCode: "ACP_SESSION_INIT_FAILED",
-            run: () => delegate.ensureSession(withAcpxSessionOptions(ensureInput)),
+            run: () => ensureDelegateSessionWithModelFallback(delegate, ensureInput),
           }),
       });
     }

extensions/brave/src/brave-web-search-provider.test.ts

@@ -43,23 +43,39 @@ afterAll(() => {
   vi.resetModules();
 });
 
+function jsonResponse(payload: unknown, init?: ResponseInit): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "content-type": "application/json" },
+    ...init,
+  });
+}
+
+function malformedJsonResponse(): Response {
+  return new Response("{ nope", {
+    status: 200,
+    headers: { "content-type": "application/json" },
+  });
+}
+
+function emptyWebSearchResponse(): Response {
+  return jsonResponse({ web: { results: [] } });
+}
+
 function installBraveLlmContextFetch() {
   const mockFetch = vi.fn(async (_input?: unknown, _init?: unknown) => {
-    return {
-      ok: true,
-      json: async () => ({
-        grounding: {
-          generic: [
-            {
-              url: "https://example.com/context",
-              title: "Context",
-              snippets: ["snippet"],
-            },
-          ],
-        },
-        sources: [],
-      }),
-    } as unknown as Response;
+    return jsonResponse({
+      grounding: {
+        generic: [
+          {
+            url: "https://example.com/context",
+            title: "Context",
+            snippets: ["snippet"],
+          },
+        ],
+      },
+      sources: [],
+    });
   });
   global.fetch = mockFetch as typeof global.fetch;
   return mockFetch;
@@ -254,10 +270,7 @@ describe("brave web search provider", () => {
   it("uses configured Brave baseUrl for web search requests", async () => {
     vi.stubEnv("BRAVE_API_KEY", "");
     const mockFetch = vi.fn(async (_input?: unknown, _init?: unknown) => {
-      return {
-        ok: true,
-        json: async () => ({ web: { results: [] } }),
-      } as unknown as Response;
+      return emptyWebSearchResponse();
     });
     global.fetch = mockFetch as typeof global.fetch;
 
@@ -310,12 +323,7 @@ describe("brave web search provider", () => {
   it("reports malformed Brave web search JSON as a provider error", async () => {
     vi.stubEnv("BRAVE_API_KEY", "");
     const mockFetch = vi.fn(async (_input?: unknown, _init?: unknown) => {
-      return {
-        ok: true,
-        json: async () => {
-          throw new SyntaxError("Unexpected token");
-        },
-      } as unknown as Response;
+      return malformedJsonResponse();
     });
     global.fetch = mockFetch as typeof global.fetch;
 
@@ -339,12 +347,7 @@ describe("brave web search provider", () => {
   it("reports malformed Brave llm-context JSON as a provider error", async () => {
     vi.stubEnv("BRAVE_API_KEY", "");
     const mockFetch = vi.fn(async (_input?: unknown, _init?: unknown) => {
-      return {
-        ok: true,
-        json: async () => {
-          throw new SyntaxError("Unexpected token");
-        },
-      } as unknown as Response;
+      return malformedJsonResponse();
     });
     global.fetch = mockFetch as typeof global.fetch;
 
@@ -428,10 +431,7 @@ describe("brave web search provider", () => {
   it("keeps Brave cache entries isolated by baseUrl", async () => {
     vi.stubEnv("BRAVE_API_KEY", "");
     const mockFetch = vi.fn(async (_input?: unknown, _init?: unknown) => {
-      return {
-        ok: true,
-        json: async () => ({ web: { results: [] } }),
-      } as unknown as Response;
+      return emptyWebSearchResponse();
     });
     global.fetch = mockFetch as typeof global.fetch;
 
@@ -573,10 +573,7 @@ describe("brave web search provider", () => {
   it("sends Brave web auth in the X-Subscription-Token header", async () => {
     vi.stubEnv("BRAVE_API_KEY", "");
     const mockFetch = vi.fn(async (_input?: unknown, _init?: unknown) => {
-      return {
-        ok: true,
-        json: async () => ({ web: { results: [] } }),
-      } as unknown as Response;
+      return emptyWebSearchResponse();
     });
     global.fetch = mockFetch as typeof global.fetch;
 
@@ -732,10 +729,7 @@ describe("brave web search provider", () => {
   it("falls back unsupported country values before calling Brave", async () => {
     vi.stubEnv("BRAVE_API_KEY", "test-key");
     const mockFetch = vi.fn(async (_input?: unknown, _init?: unknown) => {
-      return {
-        ok: true,
-        json: async () => ({ web: { results: [] } }),
-      } as unknown as Response;
+      return emptyWebSearchResponse();
     });
     global.fetch = mockFetch as typeof global.fetch;
 
@@ -763,21 +757,17 @@ describe("brave web search provider", () => {
   it("emits brave.http diagnostics for requests, responses, and cache events", async () => {
     vi.stubEnv("BRAVE_API_KEY", "");
     const mockFetch = vi.fn(async (_input?: unknown, _init?: unknown) => {
-      return {
-        ok: true,
-        status: 200,
-        json: async () => ({
-          web: {
-            results: [
-              {
-                title: "Diagnostics",
-                url: "https://example.com/diagnostics",
-                description: "debug details",
-              },
-            ],
-          },
-        }),
-      } as unknown as Response;
+      return jsonResponse({
+        web: {
+          results: [
+            {
+              title: "Diagnostics",
+              url: "https://example.com/diagnostics",
+              description: "debug details",
+            },
+          ],
+        },
+      });
     });
     global.fetch = mockFetch as typeof global.fetch;
 

extensions/chutes/implicit-provider.test.ts

@@ -15,6 +15,14 @@ function restoreEnvVar(name: string, value: string | undefined): void {
   }
 }
 
+function jsonResponse(payload: unknown, init: ResponseInit = {}): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "Content-Type": "application/json" },
+    ...init,
+  });
+}
+
 async function runChutesCatalog(params: { apiKey?: string; discoveryApiKey?: string }) {
   const provider = await registerSingleProviderPlugin(plugin);
   const result = await provider.catalog?.run({
@@ -44,10 +52,9 @@ async function withRealChutesDiscovery<T>(
   delete process.env.VITEST;
   delete process.env.NODE_ENV;
 
-  const fetchMock = vi.fn().mockResolvedValue({
-    ok: true,
-    json: async () => ({ data: [{ id: "chutes/private-model" }] }),
-  });
+  const fetchMock = vi
+    .fn()
+    .mockResolvedValue(jsonResponse({ data: [{ id: "chutes/private-model" }] }));
   globalThis.fetch = fetchMock as unknown as typeof fetch;
 
   try {

extensions/chutes/models.test.ts

@@ -15,6 +15,14 @@ function restoreEnvVar(name: string, value: string | undefined): void {
   }
 }
 
+function jsonResponse(payload: unknown, init: ResponseInit = {}): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "Content-Type": "application/json" },
+    ...init,
+  });
+}
+
 async function withLiveChutesDiscovery<T>(
   fetchMock: ReturnType<typeof vi.fn>,
   run: () => Promise<T>,
@@ -45,12 +53,11 @@ async function withLiveChutesDiscovery<T>(
 function createAuthEchoFetchMock() {
   return vi.fn().mockImplementation((_url, init?: { headers?: HeadersInit }) => {
     const auth = readAuthorizationHeader(init);
-    return Promise.resolve({
-      ok: true,
-      json: async () => ({
+    return Promise.resolve(
+      jsonResponse({
         data: [{ id: auth ? `${auth}-model` : "public-model" }],
       }),
-    });
+    );
   });
 }
 
@@ -124,9 +131,8 @@ describe("chutes-models", () => {
   });
 
   it("discoverChutesModels correctly maps API response when not in test env", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: async () => ({
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
         data: [
           { id: "zai-org/GLM-4.7-TEE" },
           {
@@ -140,7 +146,7 @@ describe("chutes-models", () => {
           { id: "new-provider/simple-model" },
         ],
       }),
-    });
+    );
     await withLiveChutesDiscovery(mockFetch, async () => {
       const models = await discoverChutesModels("test-token-real-fetch");
       expect(models.length).toBeGreaterThan(0);
@@ -158,9 +164,8 @@ describe("chutes-models", () => {
   });
 
   it("falls back from malformed live token metadata", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: async () => ({
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
         data: [
           {
             id: "provider/bad-window",
@@ -174,7 +179,7 @@ describe("chutes-models", () => {
           },
         ],
       }),
-    });
+    );
 
     await withLiveChutesDiscovery(mockFetch, async () => {
       const models = await discoverChutesModels("malformed-token-metadata");
@@ -195,14 +200,10 @@ describe("chutes-models", () => {
   it("discoverChutesModels retries without auth on 401", async () => {
     const mockFetch = vi.fn().mockImplementation((_url, init?: { headers?: HeadersInit }) => {
       if (readAuthorizationHeader(init) === "Bearer test-token-error") {
-        return Promise.resolve({
-          ok: false,
-          status: 401,
-        });
+        return Promise.resolve(new Response("", { status: 401 }));
       }
-      return Promise.resolve({
-        ok: true,
-        json: async () => ({
+      return Promise.resolve(
+        jsonResponse({
           data: [
             {
               id: "Qwen/Qwen3-32B",
@@ -232,7 +233,7 @@ describe("chutes-models", () => {
             },
           ],
         }),
-      });
+      );
     });
     await withLiveChutesDiscovery(mockFetch, async () => {
       const models = await discoverChutesModels("test-token-error");
@@ -242,10 +243,7 @@ describe("chutes-models", () => {
   });
 
   it("does not cache fallback static catalog for non-OK responses", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: false,
-      status: 503,
-    });
+    const mockFetch = vi.fn().mockResolvedValue(new Response("", { status: 503 }));
 
     await withLiveChutesDiscovery(mockFetch, async () => {
       const first = await discoverChutesModels("chutes-fallback-token");
@@ -260,27 +258,24 @@ describe("chutes-models", () => {
     const mockFetch = vi.fn().mockImplementation((_url, init?: { headers?: HeadersInit }) => {
       const auth = readAuthorizationHeader(init);
       if (auth === "Bearer chutes-token-a") {
-        return Promise.resolve({
-          ok: true,
-          json: async () => ({
+        return Promise.resolve(
+          jsonResponse({
             data: [{ id: "private/model-a" }],
           }),
-        });
+        );
       }
       if (auth === "Bearer chutes-token-b") {
-        return Promise.resolve({
-          ok: true,
-          json: async () => ({
+        return Promise.resolve(
+          jsonResponse({
             data: [{ id: "private/model-b" }],
           }),
-        });
+        );
       }
-      return Promise.resolve({
-        ok: true,
-        json: async () => ({
+      return Promise.resolve(
+        jsonResponse({
           data: [{ id: "public/model" }],
         }),
-      });
+      );
     });
     await withLiveChutesDiscovery(mockFetch, async () => {
       const modelsA = await discoverChutesModels("chutes-token-a");
@@ -325,17 +320,13 @@ describe("chutes-models", () => {
   it("does not cache 401 fallback under the failed token key", async () => {
     const mockFetch = vi.fn().mockImplementation((_url, init?: { headers?: HeadersInit }) => {
       if (readAuthorizationHeader(init) === "Bearer failed-token") {
-        return Promise.resolve({
-          ok: false,
-          status: 401,
-        });
+        return Promise.resolve(new Response("", { status: 401 }));
       }
-      return Promise.resolve({
-        ok: true,
-        json: async () => ({
+      return Promise.resolve(
+        jsonResponse({
           data: [{ id: "public/model" }],
         }),
-      });
+      );
     });
     await withLiveChutesDiscovery(mockFetch, async () => {
       await discoverChutesModels("failed-token");

extensions/codex/src/app-server/transcript-mirror.ts

@@ -360,6 +360,15 @@ export async function mirrorCodexAppServerTranscript(params: {
       sessionFile: params.sessionFile,
       ...(params.sessionKey ? { sessionKey: params.sessionKey } : {}),
       ...(params.agentId ? { agentId: params.agentId } : {}),
+      ...(params.sessionId && params.sessionKey && params.agentId
+        ? {
+            target: {
+              agentId: params.agentId,
+              sessionId: params.sessionId,
+              sessionKey: params.sessionKey,
+            },
+          }
+        : {}),
       message: update.message,
       messageId: update.messageId,
       messageSeq: update.messageSeq,

extensions/codex/src/app-server/user-input-bridge.ts

@@ -3,7 +3,12 @@
  * turns replies into app-server answer payloads.
  */
 import {
+  buildAgentHarnessUserInputAnswers,
+  deliverAgentHarnessUserInputPrompt,
   embeddedAgentLog,
+  emptyAgentHarnessUserInputAnswers,
+  type AgentHarnessUserInputOption,
+  type AgentHarnessUserInputQuestion,
   type EmbeddedRunAttemptParams,
 } from "openclaw/plugin-sdk/agent-harness-runtime";
 import { formatCodexDisplayText } from "../command-formatters.js";
@@ -19,25 +24,11 @@ type PendingUserInput = {
   threadId: string;
   turnId: string;
   itemId: string;
-  questions: UserInputQuestion[];
+  questions: AgentHarnessUserInputQuestion[];
   resolve: (value: JsonValue) => void;
   cleanup: () => void;
 };
 
-type UserInputQuestion = {
-  id: string;
-  header: string;
-  question: string;
-  isOther: boolean;
-  isSecret: boolean;
-  options: UserInputOption[] | null;
-};
-
-type UserInputOption = {
-  label: string;
-  description: string;
-};
-
 type CodexUserInputBridge = {
   handleRequest: (request: {
     id: number | string;
@@ -142,7 +133,7 @@ function readUserInputParams(value: JsonValue | undefined):
       threadId: string;
       turnId: string;
       itemId: string;
-      questions: UserInputQuestion[];
+      questions: AgentHarnessUserInputQuestion[];
     }
   | undefined {
   if (!isJsonObject(value)) {
@@ -157,11 +148,11 @@ function readUserInputParams(value: JsonValue | undefined):
   }
   const questions = questionsRaw
     .map(readQuestion)
-    .filter((question): question is UserInputQuestion => Boolean(question));
+    .filter((question): question is AgentHarnessUserInputQuestion => Boolean(question));
   return { threadId, turnId, itemId, questions };
 }
 
-function readQuestion(value: JsonValue): UserInputQuestion | undefined {
+function readQuestion(value: JsonValue): AgentHarnessUserInputQuestion | undefined {
   if (!isJsonObject(value)) {
     return undefined;
   }
@@ -181,17 +172,17 @@ function readQuestion(value: JsonValue): UserInputQuestion | undefined {
   };
 }
 
-function readOptions(value: JsonValue | undefined): UserInputOption[] | null {
+function readOptions(value: JsonValue | undefined): AgentHarnessUserInputOption[] | null {
   if (!Array.isArray(value)) {
     return null;
   }
   const options = value
     .map(readOption)
-    .filter((option): option is UserInputOption => Boolean(option));
+    .filter((option): option is AgentHarnessUserInputOption => Boolean(option));
   return options.length > 0 ? options : null;
 }
 
-function readOption(value: JsonValue): UserInputOption | undefined {
+function readOption(value: JsonValue): AgentHarnessUserInputOption | undefined {
   if (!isJsonObject(value)) {
     return undefined;
   }
@@ -202,116 +193,25 @@ function readOption(value: JsonValue): UserInputOption | undefined {
 
 async function deliverUserInputPrompt(
   params: EmbeddedRunAttemptParams,
-  questions: UserInputQuestion[],
+  questions: AgentHarnessUserInputQuestion[],
 ): Promise<void> {
-  const text = formatUserInputPrompt(questions);
-  if (params.onBlockReply) {
-    await params.onBlockReply({ text });
-    return;
-  }
-  await params.onPartialReply?.({ text });
-}
-
-function formatUserInputPrompt(questions: UserInputQuestion[]): string {
-  const lines = ["Codex needs input:"];
-  questions.forEach((question, index) => {
-    if (questions.length > 1) {
-      lines.push(
-        "",
-        `${index + 1}. ${formatCodexDisplayText(question.header)}`,
-        formatCodexDisplayText(question.question),
-      );
-    } else {
-      lines.push(
-        "",
-        formatCodexDisplayText(question.header),
-        formatCodexDisplayText(question.question),
-      );
-    }
-    if (question.isSecret) {
-      lines.push("This channel may show your reply to other participants.");
-    }
-    question.options?.forEach((option, optionIndex) => {
-      lines.push(
-        `${optionIndex + 1}. ${formatCodexDisplayText(option.label)}${
-          option.description ? ` - ${formatCodexDisplayText(option.description)}` : ""
-        }`,
-      );
-    });
-    if (question.isOther) {
-      lines.push("Other: reply with your own answer.");
-    }
+  await deliverAgentHarnessUserInputPrompt(params, questions, {
+    formatText: formatCodexDisplayText,
+    intro: "Codex needs input:",
   });
-  return lines.join("\n");
 }
 
-function buildUserInputResponse(questions: UserInputQuestion[], inputText: string): JsonObject {
+function buildUserInputResponse(
+  questions: AgentHarnessUserInputQuestion[],
+  inputText: string,
+): JsonObject {
   // Multi-question replies may use "header: answer" or numbered lines. Keep the
   // parser permissive so chat-channel replies remain ergonomic.
-  const answers: JsonObject = {};
-  if (questions.length === 1) {
-    const question = questions[0];
-    if (question) {
-      const answer = normalizeAnswer(inputText, question);
-      answers[question.id] = { answers: answer ? [answer] : [] };
-    }
-    return { answers };
-  }
-
-  const keyed = parseKeyedAnswers(inputText);
-  const fallbackLines = inputText
-    .split(/\r?\n/)
-    .map((line) => line.trim())
-    .filter(Boolean);
-  questions.forEach((question, index) => {
-    const key =
-      keyed.get(question.id.toLowerCase()) ??
-      keyed.get(question.header.toLowerCase()) ??
-      keyed.get(question.question.toLowerCase()) ??
-      keyed.get(String(index + 1));
-    const answer = key ?? fallbackLines[index] ?? "";
-    const normalized = answer ? normalizeAnswer(answer, question) : undefined;
-    answers[question.id] = { answers: normalized ? [normalized] : [] };
-  });
-  return { answers };
-}
-
-function normalizeAnswer(answer: string, question: UserInputQuestion): string | undefined {
-  const trimmed = answer.trim();
-  const options = question.options ?? [];
-  const optionIndex = /^\d+$/.test(trimmed) ? Number(trimmed) - 1 : -1;
-  const indexed = optionIndex >= 0 ? options[optionIndex] : undefined;
-  if (indexed) {
-    return indexed.label;
-  }
-  const exact = options.find((option) => option.label.toLowerCase() === trimmed.toLowerCase());
-  if (exact) {
-    return exact.label;
-  }
-  if (options.length > 0 && !question.isOther) {
-    return undefined;
-  }
-  return trimmed || undefined;
-}
-
-function parseKeyedAnswers(inputText: string): Map<string, string> {
-  const answers = new Map<string, string>();
-  for (const line of inputText.split(/\r?\n/)) {
-    const match = line.match(/^\s*([^:=-]+?)\s*[:=-]\s*(.+?)\s*$/);
-    if (!match) {
-      continue;
-    }
-    const key = match[1]?.trim().toLowerCase();
-    const value = match[2]?.trim();
-    if (key && value) {
-      answers.set(key, value);
-    }
-  }
-  return answers;
+  return buildAgentHarnessUserInputAnswers(questions, inputText) as unknown as JsonObject;
 }
 
 function emptyUserInputResponse(): JsonObject {
-  return { answers: {} };
+  return emptyAgentHarnessUserInputAnswers() as unknown as JsonObject;
 }
 
 function readString(record: JsonObject, key: string): string | undefined {

extensions/copilot/harness.test.ts

@@ -1609,6 +1609,12 @@ describe("createCopilotAgentHarness", () => {
         success: true,
         tokensRemoved: 123,
         messagesRemoved: 4,
+        summaryContent: "compacted summary",
+        contextWindow: {
+          tokenLimit: 1000,
+          currentTokens: 777,
+          messagesLength: 12,
+        },
       }));
       const disconnect = vi.fn(async () => {
         throw new Error("disconnect failed");
@@ -1649,6 +1655,7 @@ describe("createCopilotAgentHarness", () => {
         model: "gpt-4.1",
         sessionKey: "agent:main:main",
         sessionId: "oc-sess-compact-1",
+        currentTokenCount: 900,
         workspaceDir: "/this\u0000is/illegal",
         customInstructions: "Keep decisions.",
       });
@@ -1684,6 +1691,25 @@ describe("createCopilotAgentHarness", () => {
         ok: true,
         compacted: true,
         reason: "copilot-sdk-history-compacted",
+        result: {
+          summary: "compacted summary",
+          firstKeptEntryId: "",
+          tokensBefore: 900,
+          tokensAfter: 777,
+          details: {
+            success: true,
+            tokensRemoved: 123,
+            messagesRemoved: 4,
+            summaryContent: "compacted summary",
+            contextWindow: {
+              tokenLimit: 1000,
+              currentTokens: 777,
+              messagesLength: 12,
+            },
+          },
+          sessionId: "oc-sess-compact-1",
+          sessionFile: "/session.json",
+        },
       });
     });
 

extensions/copilot/harness.ts

@@ -62,6 +62,14 @@ interface CopilotHistoryCompactResult {
   tokensRemoved: number;
   messagesRemoved: number;
   summaryContent?: string;
+  contextWindow?: {
+    tokenLimit: number;
+    currentTokens: number;
+    messagesLength: number;
+    systemTokens?: number;
+    conversationTokens?: number;
+    toolDefinitionsTokens?: number;
+  };
 }
 
 interface CopilotHistoryCompactSession {
@@ -872,6 +880,21 @@ export function createCopilotAgentHarness(
         ok: true,
         compacted,
         reason: compacted ? "copilot-sdk-history-compacted" : "already under target",
+        ...(compacted
+          ? {
+              result: {
+                summary: compactResult.summaryContent ?? "",
+                firstKeptEntryId: "",
+                tokensBefore:
+                  params.currentTokenCount ??
+                  (compactResult.contextWindow?.currentTokens ?? 0) + compactResult.tokensRemoved,
+                tokensAfter: compactResult.contextWindow?.currentTokens,
+                details: compactResult,
+                sessionId: params.sessionId,
+                sessionFile: params.sessionFile,
+              },
+            }
+          : {}),
       };
     },
 

extensions/copilot/src/attempt.test.ts

@@ -3,9 +3,11 @@ import fsp from "node:fs/promises";
 import { tmpdir } from "node:os";
 import path from "node:path";
 import type { CopilotClient, Tool as SdkTool } from "@github/copilot-sdk";
-import type {
-  AgentHarnessAttemptParams,
-  AgentHarnessAttemptResult,
+import {
+  abortAgentHarnessRun,
+  queueAgentHarnessMessage,
+  type AgentHarnessAttemptParams,
+  type AgentHarnessAttemptResult,
 } from "openclaw/plugin-sdk/agent-harness-runtime";
 import type { SandboxContext } from "openclaw/plugin-sdk/agent-harness-runtime";
 import {
@@ -1171,6 +1173,36 @@ describe("runCopilotAttempt", () => {
     expect(result.externalAbort).toBe(true);
   });
 
+  it("active-run abort path marks the attempt as externally aborted", async () => {
+    const sendDeferred = createDeferred<SessionEventShape | undefined>();
+    const sessionCreated = createDeferred<FakeSession>();
+    const sdk = makeFakeSdk({
+      onCreateSession: (session) => {
+        session.sendAndWait.mockReturnValue(sendDeferred.promise);
+        session.abort.mockImplementationOnce(async () => {
+          sendDeferred.resolve(undefined);
+        });
+        sessionCreated.resolve(session);
+      },
+    });
+    const pool = makeFakePool(sdk);
+    const createToolBridge = vi.fn(async () => ({ sdkTools: [], sourceTools: [] }));
+
+    const runPromise = runCopilotAttempt(makeParams(), {
+      createToolBridge,
+      pool,
+    });
+    const session = await sessionCreated.promise;
+    await vi.waitFor(() => expect(session.sendAndWait).toHaveBeenCalledTimes(1));
+
+    expect(abortAgentHarnessRun("session-1")).toBe(true);
+    const result = await runPromise;
+
+    expect(session.abort).toHaveBeenCalledTimes(1);
+    expect(result.aborted).toBe(true);
+    expect(result.externalAbort).toBe(true);
+  });
+
   it("abort path (signal already aborted)", async () => {
     const controller = new AbortController();
     controller.abort();
@@ -1447,18 +1479,42 @@ describe("runCopilotAttempt", () => {
     expect(result.feedback).toContain("no permission policy installed");
   });
 
-  it("does not register onUserInputRequest (ask_user hidden from the model in MVP)", async () => {
-    const sdk = makeFakeSdk();
+  it("registers ask_user and resolves it from the active OpenClaw queue", async () => {
+    const onBlockReply = vi.fn();
+    const sdk = makeFakeSdk({
+      onCreateSession: (session, cfg) => {
+        session.sendAndWait.mockImplementationOnce(async () => {
+          const handler = cfg.onUserInputRequest;
+          if (typeof handler !== "function") {
+            throw new Error("expected onUserInputRequest handler");
+          }
+          const response = await handler(
+            {
+              question: "Pick a mode",
+              choices: ["Fast", "Deep"],
+              allowFreeform: false,
+            },
+            { sessionId: session.sessionId },
+          );
+          return makeAssistantMessageEvent(`selected ${response.answer}`);
+        });
+      },
+    });
     const pool = makeFakePool(sdk);
 
-    await runCopilotAttempt(makeParams(), { pool });
+    const attempt = runCopilotAttempt(makeParams({ onBlockReply }), { pool });
+
+    await vi.waitFor(() => expect(onBlockReply).toHaveBeenCalledTimes(1));
+    expect(queueAgentHarnessMessage("session-1", "2")).toBe(true);
+    const result = await attempt;
 
     const cfg = sdk.createSession.mock.calls[0]?.[0];
-    // Per the SDK contract (types.d.ts: `When provided, enables the
-    // ask_user tool allowing the agent to ask questions`), omitting the
-    // handler hides ask_user from the model entirely. The MVP keeps it
-    // hidden until a real channel/TUI prompt bridge exists.
-    expect("onUserInputRequest" in cfg).toBe(false);
+    expect(typeof cfg.onUserInputRequest).toBe("function");
+    expect(onBlockReply.mock.calls[0]?.[0]).toEqual(
+      expect.objectContaining({ text: expect.stringContaining("Pick a mode") }),
+    );
+    expect(result.assistantTexts).toEqual(["selected Deep"]);
+    expect(queueAgentHarnessMessage("session-1", "late")).toBe(false);
   });
 
   it("enableSessionTelemetry is omitted from createSession when undefined (SDK default)", async () => {
@@ -1854,6 +1910,7 @@ describe("runCopilotAttempt", () => {
   it("retains a timed-out session until later compaction reaches session.idle", async () => {
     const afterCompaction = vi.fn();
     const onDeferredCompaction = vi.fn();
+    const cleanupToolBridge = vi.fn();
     initializeGlobalHookRunner(
       createMockPluginRegistry([{ hookName: "after_compaction", handler: afterCompaction }]),
     );
@@ -1866,8 +1923,14 @@ describe("runCopilotAttempt", () => {
         );
       },
     });
+    const createToolBridge = vi.fn(async () => ({
+      cleanup: cleanupToolBridge,
+      sdkTools: [],
+      sourceTools: [],
+    }));
 
     const result = await runCopilotAttempt(makeParams(), {
+      createToolBridge,
       onDeferredCompaction,
       pool: makeFakePool(sdk),
     });
@@ -1877,6 +1940,7 @@ describe("runCopilotAttempt", () => {
     expect(onDeferredCompaction).toHaveBeenCalledWith(
       expect.objectContaining({ sdkSessionId: "sess-1" }),
     );
+    expect(cleanupToolBridge).not.toHaveBeenCalled();
     expect(activeSession?.disconnect).not.toHaveBeenCalled();
 
     activeSession?.emit("session.compaction_start", {});
@@ -1890,6 +1954,7 @@ describe("runCopilotAttempt", () => {
     await vi.waitFor(() => {
       expect(activeSession?.disconnect).toHaveBeenCalledTimes(1);
     });
+    expect(cleanupToolBridge).toHaveBeenCalledTimes(1);
   });
 
   it("does not mark a timeout after SDK compaction has completed as active compaction", async () => {
@@ -3066,7 +3131,8 @@ describe("runCopilotAttempt", () => {
   // permission policy and pollute the catalog under the default reject
   // policy. `createSessionConfig` derives `availableTools` from the
   // post-filter `sdkTools` so create- and resume-session always carry
-  // exactly the names of the tools the bridge actually exposed.
+  // exactly the names of the tools the bridge actually exposed plus the
+  // built-in `ask_user` tool owned by the registered user-input handler.
   describe("availableTools surface restriction (PR #86155 [P1] round-8)", () => {
     function makeFakeSdkTool(name: string): SdkTool {
       return {
@@ -3090,7 +3156,11 @@ describe("runCopilotAttempt", () => {
 
       await runCopilotAttempt(makeParams(), { createToolBridge, pool });
 
-      expect(readAvailableTools(sdk.createSession.mock.calls[0])).toEqual(["read", "edit"]);
+      expect(readAvailableTools(sdk.createSession.mock.calls[0])).toEqual([
+        "read",
+        "edit",
+        "builtin:ask_user",
+      ]);
     });
 
     it("forwards `[]` to the SDK when the bridge returns no tools (disable / raw / fully filtered)", async () => {
@@ -3100,12 +3170,13 @@ describe("runCopilotAttempt", () => {
       // (`modelRun: true` or `promptMode: "none"`), an empty
       // `toolsAllow: []`, and an unsupported provider to `sdkTools: []`.
       // Whatever the upstream reason, `availableTools` must be the same
-      // empty list so the SDK cannot fall back to its native catalog.
+      // ask_user-only list so the SDK cannot fall back to its native
+      // catalog while the registered user-input handler remains usable.
       const createToolBridge = vi.fn(async () => ({ sdkTools: [], sourceTools: [] }));
 
       await runCopilotAttempt(makeParams(), { createToolBridge, pool });
 
-      expect(readAvailableTools(sdk.createSession.mock.calls[0])).toEqual([]);
+      expect(readAvailableTools(sdk.createSession.mock.calls[0])).toEqual(["builtin:ask_user"]);
     });
 
     it("forwards the full bridged set when the run is unrestricted (no toolsAllow)", async () => {
@@ -3131,6 +3202,7 @@ describe("runCopilotAttempt", () => {
         "edit",
         "exec",
         "message",
+        "builtin:ask_user",
       ]);
     });
 
@@ -3156,7 +3228,7 @@ describe("runCopilotAttempt", () => {
 
       const resumeCall = sdk.resumeSession.mock.calls[0] as unknown[] | undefined;
       const resumeCfg = resumeCall?.[1] as { availableTools?: string[] };
-      expect(resumeCfg?.availableTools).toEqual(["read"]);
+      expect(resumeCfg?.availableTools).toEqual(["read", "builtin:ask_user"]);
     });
 
     it("forwards `[]` to resumeSession when the bridge returns no tools", async () => {
@@ -3175,7 +3247,7 @@ describe("runCopilotAttempt", () => {
 
       const resumeCall = sdk.resumeSession.mock.calls[0] as unknown[] | undefined;
       const resumeCfg = resumeCall?.[1] as { availableTools?: string[] };
-      expect(resumeCfg?.availableTools).toEqual([]);
+      expect(resumeCfg?.availableTools).toEqual(["builtin:ask_user"]);
     });
   });
 

extensions/copilot/src/attempt.ts

@@ -24,6 +24,8 @@ import {
   runAgentEndSideEffects,
   runAgentHarnessLlmInputHook,
   runAgentHarnessLlmOutputHook,
+  clearActiveEmbeddedRun,
+  setActiveEmbeddedRun,
 } from "openclaw/plugin-sdk/agent-harness-runtime";
 import { resolveCopilotAuth } from "./auth-bridge.js";
 import {
@@ -42,6 +44,7 @@ import {
   type SessionLike,
 } from "./event-bridge.js";
 import { createHooksBridge, type CopilotHooksConfig } from "./hooks-bridge.js";
+import { createCopilotNativeSubagentTaskMirror } from "./native-subagent-task-mirror.js";
 import {
   createPermissionBridge,
   rejectAllPolicy,
@@ -55,10 +58,12 @@ import {
 } from "./replay-shim.js";
 import type { ClientCreateOptions, CopilotClientPool, PoolKey, PooledClient } from "./runtime.js";
 import { createCopilotToolBridge } from "./tool-bridge.js";
+import { createCopilotUserInputBridge } from "./user-input-bridge.js";
 import { resolveCopilotWorkspaceBootstrapContext } from "./workspace-bootstrap.js";
 
 const SUPPORTED_PROVIDERS = new Set(["github-copilot"]);
 const BACKGROUND_COMPACTION_CANCEL_TIMEOUT_MS = 5_000;
+const COPILOT_ASK_USER_AVAILABLE_TOOLS = ["builtin:ask_user"] as const;
 
 type AttemptResultWithSdkSessionId = AgentHarnessAttemptResult & { sdkSessionId?: string };
 type PromptErrorWithCode = Error & { code?: string; cause?: unknown };
@@ -73,6 +78,7 @@ export type CopilotSessionConfig = Pick<
   | "infiniteSessions"
   | "model"
   | "onPermissionRequest"
+  | "onUserInputRequest"
   | "reasoningEffort"
   | "systemMessage"
   | "tools"
@@ -222,6 +228,8 @@ function deferBackgroundCompactionCleanup(params: {
   bridge: ReturnType<typeof attachEventBridge>;
   handle: PooledClient;
   pool: CopilotClientPool;
+  cleanupToolBridge?: () => void;
+  finalizeNativeSubagents?: () => void;
   sdkSessionId?: string;
   session: SessionLike;
   timeoutMs: number;
@@ -244,12 +252,14 @@ function deferBackgroundCompactionCleanup(params: {
         await cancelBackgroundCompactionBeforeTeardown(params.session);
         params.bridge.settleCompactionWait();
       }
+      params.finalizeNativeSubagents?.();
       params.bridge.detach();
       try {
         await params.session.disconnect();
       } catch {
         // The attempt has already returned its timeout result.
       }
+      params.cleanupToolBridge?.();
       if (outcome !== "completed" && params.sdkSessionId) {
         try {
           await params.handle.client.deleteSession(params.sdkSessionId);
@@ -403,6 +413,14 @@ export async function runCopilotAttempt(
   let handle: PooledClient | undefined;
   let session: SessionLike | undefined;
   let bridge: ReturnType<typeof attachEventBridge> | undefined;
+  const nativeSubagentTaskMirror = createCopilotNativeSubagentTaskMirror({
+    agentId: sessionAgentId,
+    now,
+    scope: input.agentHarnessTaskRuntimeScope,
+  });
+  let activeRunHandleRef: Parameters<typeof clearActiveEmbeddedRun>[1] | undefined;
+  let userInputBridgeRef: ReturnType<typeof createCopilotUserInputBridge> | undefined;
+  let cleanupToolBridge: (() => void) | undefined;
   let releaseError: Error | undefined;
   let downgradedFromResume = false;
   let resumeFailureRecovered = false;
@@ -415,16 +433,24 @@ export async function runCopilotAttempt(
   // `src/agents/pi-embedded-runner/run/types.ts:139`.
   let yieldDetected = false;
 
-  const onAbort = () => {
+  const markExternalAbort = () => {
     abortRequested = true;
     externalAbort = true;
     aborted = true;
+  };
+
+  const abortActiveSession = () => {
+    markExternalAbort();
     if (settled || !sentTurnStarted || !session) {
       return;
     }
     void session.abort().catch(() => undefined);
   };
 
+  const onAbort = () => {
+    abortActiveSession();
+  };
+
   params.abortSignal?.addEventListener("abort", onAbort, { once: true });
 
   // Sandbox parity with PI (`src/agents/pi-embedded-runner/run/attempt.ts:1232-1244`):
@@ -575,6 +601,7 @@ export async function runCopilotAttempt(
             startedAt,
           }),
       });
+      cleanupToolBridge = toolBridge.cleanup;
       sdkTools = toolBridge.sdkTools;
     } catch (error: unknown) {
       const result = createResult(input, {
@@ -655,6 +682,11 @@ export async function runCopilotAttempt(
       });
     };
     const hasNativePromptHook = Boolean(attemptInput.hooksConfig?.onUserPromptSubmitted);
+    const userInputBridge = createCopilotUserInputBridge({
+      paramsForRun: attemptInput,
+      signal: params.abortSignal,
+    });
+    userInputBridgeRef = userInputBridge;
     const sessionConfig = createSessionConfig(
       attemptInput,
       modelRef.id,
@@ -663,6 +695,7 @@ export async function runCopilotAttempt(
       promptBuild.developerInstructions || undefined,
       effectiveWorkspaceDir,
       effectiveCwd,
+      userInputBridge.onUserInputRequest,
       hasNativePromptHook
         ? {
             onUserPromptSubmitted: ({ additionalContext, prompt }) =>
@@ -723,6 +756,8 @@ export async function runCopilotAttempt(
     }
     bridge = attachEventBridge(session, {
       onAssistantDelta: input.onAssistantDelta,
+      onAgentEvent: input.onAgentEvent,
+      onNativeSubagentEvent: (event) => nativeSubagentTaskMirror?.handleEvent(event),
       onCompactionStart: async () => {
         const sessionFile = readString(input.sessionFile);
         if (!sessionFile) {
@@ -748,6 +783,29 @@ export async function runCopilotAttempt(
       isAborted: () => aborted,
     });
 
+    const activeRunHandle = {
+      kind: "embedded" as const,
+      queueMessage: async (text: string) => {
+        if (userInputBridge.handleQueuedMessage(text)) {
+          return;
+        }
+        throw new Error("Copilot runtime is not waiting for user input.");
+      },
+      isStreaming: () => !settled && !aborted,
+      isCompacting: () => bridge?.isCompacting() ?? false,
+      sourceReplyDeliveryMode: input.sourceReplyDeliveryMode,
+      cancel: () => {
+        userInputBridge.cancelPending();
+        abortActiveSession();
+      },
+      abort: () => {
+        userInputBridge.cancelPending();
+        abortActiveSession();
+      },
+    };
+    setActiveEmbeddedRun(input.sessionId, activeRunHandle, input.sessionKey, input.sessionFile);
+    activeRunHandleRef = activeRunHandle;
+
     const messageOptions = await createMessageOptions(attemptInput, {
       effectiveCwd,
       effectiveWorkspaceDir,
@@ -765,6 +823,7 @@ export async function runCopilotAttempt(
       }
       const result = await session.sendAndWait(messageOptions, input.timeoutMs);
       await bridge.awaitDeltaChain();
+      await bridge.awaitAgentEventChain();
       if (!bridge.recordSendResult(result) && !aborted) {
         // SDK sendAndWait returning undefined is treated as a timeout by the
         // capability inventory. Do not call session.abort() here: OpenClaw may
@@ -800,12 +859,22 @@ export async function runCopilotAttempt(
         } catch {
           // delta-flush failure must not mask the timeout state
         }
+        await bridge?.awaitAgentEventChain();
       } else {
         promptError = toError(error);
       }
     }
   } finally {
     settled = true;
+    userInputBridgeRef?.cancelPending();
+    if (activeRunHandleRef) {
+      clearActiveEmbeddedRun(
+        input.sessionId,
+        activeRunHandleRef,
+        input.sessionKey,
+        input.sessionFile,
+      );
+    }
     const retainSessionForDeferredCleanup =
       bridge?.hasObservedCompaction() || (timedOut && bridge?.hasObservedSessionIdle() === false);
     if (retainSessionForDeferredCleanup && bridge && session && handle) {
@@ -820,6 +889,8 @@ export async function runCopilotAttempt(
         abortSignal: cleanupAbort.signal,
         awaitSessionIdle: !bridge.hasObservedSessionIdle(),
         bridge,
+        cleanupToolBridge,
+        finalizeNativeSubagents: () => nativeSubagentTaskMirror?.finalizeActiveRuns(),
         handle,
         pool: deps.pool,
         sdkSessionId,
@@ -848,6 +919,9 @@ export async function runCopilotAttempt(
       // defines as no background agents in flight. Timeouts retain the bridge
       // until that event so compaction that starts after the timer still completes.
       await bridge?.awaitCompactionChain();
+      await bridge?.awaitAgentEventChain();
+      nativeSubagentTaskMirror?.finalizeActiveRuns();
+      cleanupToolBridge?.();
       bridge?.detach();
       params.abortSignal?.removeEventListener("abort", onAbort);
 
@@ -959,6 +1033,7 @@ export async function runCopilotAttempt(
     await dualWriteCopilotTranscriptBestEffort({
       sessionFile: sessionFileForMirror,
       sessionKey: readString((input as { sessionKey?: unknown }).sessionKey),
+      sessionId: readString(input.sessionId),
       agentId: readString(input.agentId),
       messages: taggedMessages,
       idempotencyScope: sessionIdForScope ? `copilot:${sessionIdForScope}` : undefined,
@@ -1118,6 +1193,7 @@ function createSessionConfig(
   systemMessageContent: string | undefined,
   effectiveWorkspaceDir: string | undefined,
   effectiveCwd: string | undefined,
+  onUserInputRequest: NonNullable<SessionConfig["onUserInputRequest"]>,
   hooksBridgeOptions?: Parameters<typeof createHooksBridge>[1],
 ): CopilotSessionConfig {
   const permissionPolicy = params.permissionPolicy ?? rejectAllPolicy;
@@ -1145,10 +1221,9 @@ function createSessionConfig(
     // tool wrapper, and the SDK gate is a safety net for kinds we
     // don't surface. See permission-bridge.ts and docs/plugins/copilot.md.
     onPermissionRequest: createPermissionBridge(permissionPolicy),
-    // `onUserInputRequest` is intentionally NOT registered: per the SDK
-    // contract, omitting the handler hides the `ask_user` tool from the
-    // model entirely. Interactive ask_user will need a real channel/TUI
-    // prompt bridge before this runtime can expose the handler.
+    // Registers the SDK ask_user bridge. The bridge itself owns pending
+    // reply routing so generic mid-run steering still fails closed.
+    onUserInputRequest,
     // Preserve the shipped native SDK hook contract. These callbacks expose
     // Copilot-specific events and decisions that generic lifecycle hooks do
     // not model.
@@ -1166,8 +1241,9 @@ function createSessionConfig(
     ...(infiniteSessions ? { infiniteSessions } : {}),
     reasoningEffort: params.reasoningEffort,
     tools: sdkTools,
-    // Restrict the SDK's tool catalog to exactly the bridged tool names
-    // returned by `createCopilotToolBridge`. Without this, the SDK
+    // Restrict the SDK's tool catalog to the bridged tool names returned
+    // by `createCopilotToolBridge` plus the built-in `ask_user` tool owned
+    // by `onUserInputRequest`. Without this, the SDK
     // would still expose its native read/write/shell/url/mcp/memory/
     // hook tools to the model alongside our overrides, which would
     // bypass OpenClaw's wrapped-tool enforcement under any permissive
@@ -1182,7 +1258,7 @@ function createSessionConfig(
     // `@github/copilot-sdk/dist/types.d.ts:1198` (it picks
     // `availableTools`, so the spread into `resumeSession` covers
     // the resume path too).
-    availableTools: sdkTools.map((tool) => tool.name),
+    availableTools: buildCopilotAvailableTools(sdkTools),
     workingDirectory:
       effectiveCwd ?? effectiveWorkspaceDir ?? readResolvedAttemptPath(params.workspaceDir),
     // When a task runs from a sub-cwd, keep SDK-native project docs
@@ -1228,6 +1304,10 @@ function createSessionConfig(
   };
 }
 
+function buildCopilotAvailableTools(sdkTools: SdkTool[]): string[] {
+  return [...new Set([...sdkTools.map((tool) => tool.name), ...COPILOT_ASK_USER_AVAILABLE_TOOLS])];
+}
+
 async function createMessageOptions(
   params: AttemptParamsLike,
   context: {

extensions/copilot/src/dual-write-transcripts.ts

@@ -96,6 +96,7 @@ function buildMirrorDedupeIdentity(message: MirroredAgentMessage): string {
 export interface MirrorCopilotTranscriptParams {
   sessionFile: string;
   sessionKey?: string;
+  sessionId?: string;
   agentId?: string;
   messages: AgentMessage[];
   /**
@@ -168,7 +169,20 @@ export async function mirrorCopilotTranscript(
   }
 
   if (params.sessionKey) {
-    emitSessionTranscriptUpdate({ sessionFile: params.sessionFile, sessionKey: params.sessionKey });
+    emitSessionTranscriptUpdate({
+      sessionFile: params.sessionFile,
+      sessionKey: params.sessionKey,
+      ...(params.agentId ? { agentId: params.agentId } : {}),
+      ...(params.sessionId && params.agentId
+        ? {
+            target: {
+              agentId: params.agentId,
+              sessionId: params.sessionId,
+              sessionKey: params.sessionKey,
+            },
+          }
+        : {}),
+    });
   } else {
     emitSessionTranscriptUpdate(params.sessionFile);
   }

extensions/copilot/src/event-bridge.test.ts

@@ -15,6 +15,11 @@ const REGISTERED_EVENT_TYPES = [
   "assistant.usage",
   "tool.execution_start",
   "tool.execution_complete",
+  "session.plan_changed",
+  "exit_plan_mode.requested",
+  "subagent.started",
+  "subagent.completed",
+  "subagent.failed",
   "session.compaction_start",
   "session.compaction_complete",
   "session.idle",
@@ -455,6 +460,78 @@ describe("attachEventBridge", () => {
     });
   });
 
+  it("projects Copilot plan events through the generic plan stream", async () => {
+    const session = createFakeSession();
+    const onAgentEvent = vi.fn().mockResolvedValue(undefined);
+    const bridge = attachEventBridge(session, {
+      getSdkSessionId: () => "sdk-session-id",
+      isAborted: () => false,
+      onAgentEvent,
+    });
+
+    session.emit(
+      "session.plan_changed",
+      makeEvent("session.plan_changed", { operation: "update" }),
+    );
+    session.emit(
+      "exit_plan_mode.requested",
+      makeEvent("exit_plan_mode.requested", {
+        actions: ["approve", "edit"],
+        planContent: "# Plan\n- inspect\n- patch",
+        recommendedAction: "approve",
+        requestId: "request-1",
+        summary: "Plan ready",
+      }),
+    );
+
+    await bridge.awaitAgentEventChain();
+
+    expect(onAgentEvent).toHaveBeenCalledTimes(2);
+    expect(onAgentEvent).toHaveBeenNthCalledWith(1, {
+      stream: "plan",
+      data: {
+        phase: "update",
+        title: "Plan updated",
+        source: "copilot-sdk",
+        operation: "update",
+      },
+    });
+    expect(onAgentEvent).toHaveBeenNthCalledWith(2, {
+      stream: "plan",
+      data: {
+        phase: "update",
+        title: "Plan updated",
+        source: "copilot-sdk",
+        explanation: "Plan ready",
+        steps: ["# Plan", "inspect", "patch"],
+        actions: ["approve", "edit"],
+        requestId: "request-1",
+        recommendedAction: "approve",
+      },
+    });
+  });
+
+  it("forwards native Copilot subagent lifecycle events to the adapter", () => {
+    const session = createFakeSession();
+    const onNativeSubagentEvent = vi.fn();
+    const bridge = attachEventBridge(session, {
+      getSdkSessionId: () => "sdk-session-id",
+      isAborted: () => false,
+      onNativeSubagentEvent,
+    });
+    const event = makeEvent("subagent.started", {
+      agentDescription: "inspect the repository",
+      agentDisplayName: "Researcher",
+      agentName: "researcher",
+      toolCallId: "call-1",
+    });
+
+    session.emit("subagent.started", event);
+
+    expect(onNativeSubagentEvent).toHaveBeenCalledWith(event);
+    bridge.detach();
+  });
+
   it("preserves all-zero usage snapshot after an invalid assistant.usage event", () => {
     const session = createFakeSession();
     const bridge = attachEventBridge(session, {

extensions/copilot/src/event-bridge.ts

@@ -41,6 +41,16 @@ export interface SessionLike {
 
 export interface EventBridgeOptions {
   onAssistantDelta?: (payload: OnAssistantDeltaPayload) => void | Promise<void>;
+  onAgentEvent?: (event: {
+    stream: "item" | "plan";
+    data: Record<string, unknown>;
+  }) => void | Promise<void>;
+  onNativeSubagentEvent?: (
+    event: Extract<
+      SessionEvent,
+      { type: "subagent.started" | "subagent.completed" | "subagent.failed" }
+    >,
+  ) => void;
   onCompactionComplete?: (payload: {
     messagesRemoved?: number;
     success: boolean;
@@ -72,6 +82,7 @@ export interface EventBridgeController {
   awaitSessionIdle(): Promise<void>;
   settleCompactionWait(): void;
   awaitDeltaChain(): Promise<void>;
+  awaitAgentEventChain(): Promise<void>;
   hasObservedCompaction(): boolean;
   hasObservedSessionIdle(): boolean;
   isCompacting(): boolean;
@@ -103,6 +114,7 @@ export function attachEventBridge(
   let observedCompaction = false;
   let deltaQueue = Promise.resolve();
   let deltaChain = Promise.resolve();
+  let agentEventChain = Promise.resolve();
   let compactionChain = Promise.resolve();
   let compactionIdle = Promise.resolve();
   let resolveCompactionIdle: (() => void) | undefined;
@@ -191,6 +203,51 @@ export function attachEventBridge(
     }
   });
 
+  registerListener(session, unsubscribeFns, "session.plan_changed", (event) => {
+    enqueueAgentEvent({
+      stream: "plan",
+      data: {
+        phase: "update",
+        title: "Plan updated",
+        source: "copilot-sdk",
+        operation: event.data.operation,
+        ...(event.agentId ? { agentId: event.agentId } : {}),
+      },
+    });
+  });
+
+  registerListener(session, unsubscribeFns, "exit_plan_mode.requested", (event) => {
+    const steps = splitPlanText(event.data.planContent);
+    enqueueAgentEvent({
+      stream: "plan",
+      data: {
+        phase: "update",
+        title: "Plan updated",
+        source: "copilot-sdk",
+        ...(event.data.summary ? { explanation: event.data.summary } : {}),
+        ...(steps.length > 0 ? { steps } : {}),
+        ...(event.data.actions.length > 0 ? { actions: event.data.actions } : {}),
+        ...(event.data.requestId ? { requestId: event.data.requestId } : {}),
+        ...(event.data.recommendedAction
+          ? { recommendedAction: event.data.recommendedAction }
+          : {}),
+        ...(event.agentId ? { agentId: event.agentId } : {}),
+      },
+    });
+  });
+
+  registerListener(session, unsubscribeFns, "subagent.started", (event) => {
+    forwardNativeSubagentEvent(event);
+  });
+
+  registerListener(session, unsubscribeFns, "subagent.completed", (event) => {
+    forwardNativeSubagentEvent(event);
+  });
+
+  registerListener(session, unsubscribeFns, "subagent.failed", (event) => {
+    forwardNativeSubagentEvent(event);
+  });
+
   registerListener(session, unsubscribeFns, "session.compaction_start", (event) => {
     if (!isRootCompactionEvent(event)) {
       return;
@@ -276,6 +333,9 @@ export function attachEventBridge(
     awaitDeltaChain() {
       return deltaChain;
     },
+    awaitAgentEventChain() {
+      return agentEventChain;
+    },
     hasObservedCompaction() {
       return observedCompaction;
     },
@@ -334,6 +394,31 @@ export function attachEventBridge(
     compactionChain = queued.catch(() => undefined);
   }
 
+  function enqueueAgentEvent(event: {
+    stream: "item" | "plan";
+    data: Record<string, unknown>;
+  }): void {
+    const callback = options.onAgentEvent;
+    if (!callback) {
+      return;
+    }
+    const invoke = () => callback(event);
+    agentEventChain = agentEventChain.then(invoke, invoke).catch(() => undefined);
+  }
+
+  function forwardNativeSubagentEvent(
+    event: Extract<
+      SessionEvent,
+      { type: "subagent.started" | "subagent.completed" | "subagent.failed" }
+    >,
+  ): void {
+    try {
+      options.onNativeSubagentEvent?.(event);
+    } catch {
+      // Native task mirroring must not corrupt the Copilot turn.
+    }
+  }
+
   async function awaitStableCompaction(): Promise<void> {
     const idle = activeCompactionCount > 0 ? compactionIdle : undefined;
     if (idle) {
@@ -456,6 +541,13 @@ function joinReasoning(order: string[], reasoningById: Map<string, string>): str
   return order.map((reasoningId) => reasoningById.get(reasoningId) ?? "").join("");
 }
 
+function splitPlanText(text: string | undefined): string[] {
+  return (text ?? "")
+    .split(/\r?\n/)
+    .map((line) => line.trim().replace(/^[-*]\s+/, ""))
+    .filter((line) => line.length > 0);
+}
+
 function readString(value: unknown): string | undefined {
   return typeof value === "string" && value.length > 0 ? value : undefined;
 }

extensions/copilot/src/native-subagent-task-mirror.test.ts

@@ -0,0 +1,200 @@
+import type { SessionEvent } from "@github/copilot-sdk";
+import type {
+  AgentHarnessTaskRecord,
+  AgentHarnessTaskRuntime,
+} from "openclaw/plugin-sdk/agent-harness-task-runtime";
+import { describe, expect, it, vi } from "vitest";
+import {
+  CopilotNativeSubagentTaskMirror,
+  createCopilotNativeSubagentTaskMirror,
+} from "./native-subagent-task-mirror.js";
+
+type NativeSubagentEventType = "subagent.started" | "subagent.completed" | "subagent.failed";
+
+function makeEvent<T extends NativeSubagentEventType>(
+  type: T,
+  data: Extract<SessionEvent, { type: T }>["data"],
+  agentId?: string,
+): Extract<SessionEvent, { type: T }> {
+  return {
+    data,
+    id: `${type}-id`,
+    parentId: null,
+    timestamp: "2024-01-01T00:00:00.000Z",
+    type,
+    ...(agentId ? { agentId } : {}),
+  } as Extract<SessionEvent, { type: T }>;
+}
+
+function createRuntime() {
+  const task = {} as AgentHarnessTaskRecord;
+  return {
+    tryCreateRunningTaskRun: vi.fn(() => task),
+    recordTaskRunProgressByRunId: vi.fn(() => []),
+    finalizeTaskRunByRunId: vi.fn(() => []),
+  } satisfies Pick<
+    AgentHarnessTaskRuntime,
+    "tryCreateRunningTaskRun" | "recordTaskRunProgressByRunId" | "finalizeTaskRunByRunId"
+  >;
+}
+
+describe("CopilotNativeSubagentTaskMirror", () => {
+  it("does not create a mirror without a host-issued task scope", () => {
+    expect(createCopilotNativeSubagentTaskMirror({})).toBeUndefined();
+  });
+
+  it("mirrors start and completion using agentId with toolCallId fallback", () => {
+    const runtime = createRuntime();
+    const mirror = new CopilotNativeSubagentTaskMirror(
+      { agentId: "parent-agent", now: () => 100 },
+      runtime,
+    );
+
+    mirror.handleEvent(
+      makeEvent(
+        "subagent.started",
+        {
+          agentDescription: "inspect the repository",
+          agentDisplayName: "Researcher",
+          agentName: "researcher",
+          toolCallId: "call-1",
+        },
+        "child-1",
+      ),
+    );
+    mirror.handleEvent(
+      makeEvent(
+        "subagent.completed",
+        {
+          agentDisplayName: "Researcher",
+          agentName: "researcher",
+          toolCallId: "call-1",
+          totalToolCalls: 2,
+          totalTokens: 30,
+        },
+        "child-1",
+      ),
+    );
+
+    expect(runtime.tryCreateRunningTaskRun).toHaveBeenCalledWith({
+      sourceId: "call-1",
+      agentId: "parent-agent",
+      runId: "copilot-agent:child-1",
+      label: "Researcher",
+      task: "inspect the repository",
+      notifyPolicy: "silent",
+      deliveryStatus: "not_applicable",
+      preferMetadata: true,
+      startedAt: 100,
+      lastEventAt: 100,
+      progressSummary: "Copilot native subagent started.",
+    });
+    expect(runtime.finalizeTaskRunByRunId).toHaveBeenCalledWith({
+      runId: "copilot-agent:child-1",
+      status: "succeeded",
+      endedAt: 100,
+      lastEventAt: 100,
+      progressSummary: "Copilot native subagent completed.",
+      terminalSummary: "Copilot native subagent completed (2 tool calls, 30 tokens).",
+    });
+  });
+
+  it("uses toolCallId when the SDK omits agentId", () => {
+    const runtime = createRuntime();
+    const mirror = new CopilotNativeSubagentTaskMirror({ now: () => 200 }, runtime);
+
+    mirror.handleEvent(
+      makeEvent("subagent.started", {
+        agentDescription: "",
+        agentDisplayName: "Researcher",
+        agentName: "researcher",
+        toolCallId: "call-2",
+      }),
+    );
+    mirror.handleEvent(
+      makeEvent("subagent.failed", {
+        agentDisplayName: "Researcher",
+        agentName: "researcher",
+        error: "failed",
+        toolCallId: "call-2",
+      }),
+    );
+
+    expect(runtime.finalizeTaskRunByRunId).toHaveBeenCalledWith(
+      expect.objectContaining({
+        runId: "copilot-agent:call-2",
+        status: "failed",
+        error: "failed",
+      }),
+    );
+  });
+
+  it("keeps parallel subagents distinct when they share a parent tool call", () => {
+    const runtime = createRuntime();
+    const mirror = new CopilotNativeSubagentTaskMirror({ now: () => 250 }, runtime);
+
+    for (const agentId of ["child-1", "child-2"]) {
+      mirror.handleEvent(
+        makeEvent(
+          "subagent.started",
+          {
+            agentDescription: `inspect ${agentId}`,
+            agentDisplayName: "Researcher",
+            agentName: "researcher",
+            toolCallId: "call-shared",
+          },
+          agentId,
+        ),
+      );
+    }
+    for (const agentId of ["child-1", "child-2"]) {
+      mirror.handleEvent(
+        makeEvent(
+          "subagent.completed",
+          {
+            agentDisplayName: "Researcher",
+            agentName: "researcher",
+            toolCallId: "call-shared",
+          },
+          agentId,
+        ),
+      );
+    }
+
+    expect(runtime.tryCreateRunningTaskRun).toHaveBeenCalledTimes(2);
+    expect(runtime.finalizeTaskRunByRunId).toHaveBeenCalledTimes(2);
+    expect(runtime.finalizeTaskRunByRunId).toHaveBeenNthCalledWith(
+      1,
+      expect.objectContaining({ runId: "copilot-agent:child-1" }),
+    );
+    expect(runtime.finalizeTaskRunByRunId).toHaveBeenNthCalledWith(
+      2,
+      expect.objectContaining({ runId: "copilot-agent:child-2" }),
+    );
+  });
+
+  it("finalizes active tasks when the parent attempt tears down", () => {
+    const runtime = createRuntime();
+    const mirror = new CopilotNativeSubagentTaskMirror({ now: () => 300 }, runtime);
+
+    mirror.handleEvent(
+      makeEvent("subagent.started", {
+        agentDescription: "inspect",
+        agentDisplayName: "Researcher",
+        agentName: "researcher",
+        toolCallId: "call-3",
+      }),
+    );
+    mirror.finalizeActiveRuns();
+
+    expect(runtime.finalizeTaskRunByRunId).toHaveBeenCalledWith({
+      runId: "copilot-agent:call-3",
+      status: "cancelled",
+      endedAt: 300,
+      lastEventAt: 300,
+      error: "Copilot native subagent ended with its parent attempt.",
+      progressSummary: "Copilot native subagent cancelled with its parent attempt.",
+      terminalSummary: "Copilot native subagent cancelled.",
+    });
+  });
+});

extensions/copilot/src/native-subagent-task-mirror.ts

@@ -0,0 +1,199 @@
+import type { SessionEvent } from "@github/copilot-sdk";
+import {
+  createAgentHarnessTaskRuntime,
+  type AgentHarnessTaskRuntime,
+  type AgentHarnessTaskRuntimeScope,
+} from "openclaw/plugin-sdk/agent-harness-task-runtime";
+
+const COPILOT_NATIVE_SUBAGENT_TASK_KIND = "copilot-native";
+const COPILOT_NATIVE_SUBAGENT_RUN_ID_PREFIX = "copilot-agent:";
+
+type CopilotNativeSubagentEvent = Extract<
+  SessionEvent,
+  { type: "subagent.started" | "subagent.completed" | "subagent.failed" }
+>;
+
+type TaskLifecycleRuntime = Pick<
+  AgentHarnessTaskRuntime,
+  "tryCreateRunningTaskRun" | "recordTaskRunProgressByRunId" | "finalizeTaskRunByRunId"
+>;
+
+export function createCopilotNativeSubagentTaskMirror(params: {
+  agentId?: string;
+  now?: () => number;
+  scope?: AgentHarnessTaskRuntimeScope;
+}): CopilotNativeSubagentTaskMirror | undefined {
+  if (!params.scope) {
+    return undefined;
+  }
+  return new CopilotNativeSubagentTaskMirror(
+    {
+      agentId: params.agentId,
+      now: params.now,
+    },
+    createAgentHarnessTaskRuntime({
+      runtime: "subagent",
+      taskKind: COPILOT_NATIVE_SUBAGENT_TASK_KIND,
+      scope: params.scope,
+      runIdPrefix: COPILOT_NATIVE_SUBAGENT_RUN_ID_PREFIX,
+    }),
+  );
+}
+
+export class CopilotNativeSubagentTaskMirror {
+  private readonly runIdByAgentId = new Map<string, string>();
+  private readonly runIdByToolCallId = new Map<string, string>();
+  private readonly terminalRunIds = new Set<string>();
+  private readonly activeRunIds = new Set<string>();
+  private readonly now: () => number;
+
+  constructor(
+    private readonly params: { agentId?: string; now?: () => number },
+    private readonly runtime: TaskLifecycleRuntime,
+  ) {
+    this.now = params.now ?? Date.now;
+  }
+
+  handleEvent(event: CopilotNativeSubagentEvent): void {
+    const toolCallId = event.data.toolCallId.trim();
+    if (!toolCallId) {
+      return;
+    }
+    const runId = this.resolveRunId(event);
+    if (event.type === "subagent.started") {
+      this.handleStarted(event, runId, toolCallId);
+      return;
+    }
+    if (event.type === "subagent.completed") {
+      this.handleCompleted(event, runId);
+      return;
+    }
+    this.handleFailed(event, runId);
+  }
+
+  finalizeActiveRuns(): void {
+    const eventAt = this.now();
+    for (const runId of this.activeRunIds) {
+      this.terminalRunIds.add(runId);
+      this.runtime.finalizeTaskRunByRunId({
+        runId,
+        status: "cancelled",
+        endedAt: eventAt,
+        lastEventAt: eventAt,
+        error: "Copilot native subagent ended with its parent attempt.",
+        progressSummary: "Copilot native subagent cancelled with its parent attempt.",
+        terminalSummary: "Copilot native subagent cancelled.",
+      });
+    }
+    this.activeRunIds.clear();
+  }
+
+  private handleStarted(
+    event: Extract<CopilotNativeSubagentEvent, { type: "subagent.started" }>,
+    runId: string,
+    toolCallId: string,
+  ): void {
+    const agentId = event.agentId?.trim();
+    const existingRunId = agentId
+      ? this.runIdByAgentId.get(agentId)
+      : this.runIdByToolCallId.get(toolCallId);
+    if (existingRunId) {
+      return;
+    }
+    const eventAt = this.now();
+    const label = event.data.agentDisplayName.trim() || event.data.agentName.trim();
+    const task = event.data.agentDescription.trim() || `Copilot native subagent ${label}`;
+    const taskRecord = this.runtime.tryCreateRunningTaskRun({
+      sourceId: toolCallId,
+      agentId: this.params.agentId,
+      runId,
+      label: label || "Copilot subagent",
+      task,
+      notifyPolicy: "silent",
+      deliveryStatus: "not_applicable",
+      preferMetadata: true,
+      startedAt: eventAt,
+      lastEventAt: eventAt,
+      progressSummary: "Copilot native subagent started.",
+    });
+    if (!taskRecord) {
+      return;
+    }
+    if (agentId) {
+      this.runIdByAgentId.set(agentId, runId);
+    } else {
+      this.runIdByToolCallId.set(toolCallId, runId);
+    }
+    this.terminalRunIds.delete(runId);
+    this.activeRunIds.add(runId);
+  }
+
+  private handleCompleted(
+    event: Extract<CopilotNativeSubagentEvent, { type: "subagent.completed" }>,
+    runId: string,
+  ): void {
+    if (this.terminalRunIds.has(runId)) {
+      return;
+    }
+    const eventAt = this.now();
+    this.terminalRunIds.add(runId);
+    this.activeRunIds.delete(runId);
+    this.runtime.finalizeTaskRunByRunId({
+      runId,
+      status: "succeeded",
+      endedAt: eventAt,
+      lastEventAt: eventAt,
+      progressSummary: "Copilot native subagent completed.",
+      terminalSummary: buildCompletionSummary(event),
+    });
+  }
+
+  private handleFailed(
+    event: Extract<CopilotNativeSubagentEvent, { type: "subagent.failed" }>,
+    runId: string,
+  ): void {
+    if (this.terminalRunIds.has(runId)) {
+      return;
+    }
+    const eventAt = this.now();
+    this.terminalRunIds.add(runId);
+    this.activeRunIds.delete(runId);
+    this.runtime.finalizeTaskRunByRunId({
+      runId,
+      status: "failed",
+      endedAt: eventAt,
+      lastEventAt: eventAt,
+      error: event.data.error,
+      progressSummary: "Copilot native subagent failed.",
+      terminalSummary: "Copilot native subagent failed.",
+    });
+  }
+
+  private resolveRunId(event: CopilotNativeSubagentEvent): string {
+    const agentId = event.agentId?.trim();
+    if (agentId) {
+      const existing = this.runIdByAgentId.get(agentId);
+      if (existing) {
+        return existing;
+      }
+    }
+    const existing = this.runIdByToolCallId.get(event.data.toolCallId);
+    if (existing) {
+      return existing;
+    }
+    const identity = agentId || event.data.toolCallId.trim();
+    return `${COPILOT_NATIVE_SUBAGENT_RUN_ID_PREFIX}${identity}`;
+  }
+}
+
+function buildCompletionSummary(
+  event: Extract<CopilotNativeSubagentEvent, { type: "subagent.completed" }>,
+): string {
+  const details = [
+    event.data.totalToolCalls !== undefined ? `${event.data.totalToolCalls} tool calls` : undefined,
+    event.data.totalTokens !== undefined ? `${event.data.totalTokens} tokens` : undefined,
+  ].filter((value): value is string => value !== undefined);
+  return details.length > 0
+    ? `Copilot native subagent completed (${details.join(", ")}).`
+    : "Copilot native subagent completed.";
+}

extensions/copilot/src/tool-bridge.test.ts

@@ -156,11 +156,184 @@ describe("createCopilotToolBridge", () => {
       sessionId: "session-1",
     });
 
-    expect(result.sourceTools).toBe(sourceTools);
+    expect(result.sourceTools).toEqual(sourceTools);
     expect(result.sdkTools).toHaveLength(2);
     expect(result.sdkTools.map((tool) => tool.name)).toEqual(["tool-a", "tool-b"]);
   });
 
+  it("compacts the Copilot tool surface behind tool_search controls when enabled", async () => {
+    const createOpenClawCodingTools = vi.fn(async (opts: unknown) => {
+      const includeToolSearchControls = Boolean(
+        (opts as { includeToolSearchControls?: boolean }).includeToolSearchControls,
+      );
+      return includeToolSearchControls
+        ? [
+            makeTool({ name: "tool_search_code" }),
+            makeTool({ name: "fake_hidden" }),
+            makeTool({ name: "read" }),
+          ]
+        : [makeTool({ name: "fake_hidden" }), makeTool({ name: "read" })];
+    });
+
+    const result = await createCopilotToolBridge({
+      agentId: "agent-1",
+      attemptParams: {
+        config: { tools: { toolSearch: true } },
+        runId: "run-tool-search",
+        sessionKey: "agent:main:main",
+      } as never,
+      createOpenClawCodingTools,
+      modelId: "gpt-4o",
+      modelProvider: "github-copilot",
+      sessionId: "session-1",
+    });
+
+    expect(createOpenClawCodingTools).toHaveBeenCalledWith(
+      expect.objectContaining({
+        includeToolSearchControls: true,
+        toolSearchCatalogRef: expect.any(Object),
+        toolSearchCatalogExecutor: expect.any(Function),
+      }),
+    );
+    expect(result.sourceTools.map((tool) => tool.name)).toEqual(["tool_search_code"]);
+    expect(result.sdkTools.map((tool) => tool.name)).toEqual(["tool_search_code"]);
+  });
+
+  it("keeps tool_search controls visible when a narrow allowlist is active", async () => {
+    const createOpenClawCodingTools = vi.fn(async (opts: unknown) => {
+      const includeToolSearchControls = Boolean(
+        (opts as { includeToolSearchControls?: boolean }).includeToolSearchControls,
+      );
+      return includeToolSearchControls
+        ? [makeTool({ name: "tool_search_code" }), makeTool({ name: "read" })]
+        : [makeTool({ name: "read" })];
+    });
+
+    const result = await createCopilotToolBridge({
+      agentId: "agent-1",
+      attemptParams: {
+        config: { tools: { toolSearch: true } },
+        runId: "run-tool-search",
+        sessionKey: "agent:main:main",
+        toolsAllow: ["read"],
+      } as never,
+      createOpenClawCodingTools,
+      modelId: "gpt-4o",
+      modelProvider: "github-copilot",
+      sessionId: "session-1",
+    });
+
+    expect(result.sourceTools.map((tool) => tool.name)).toEqual(["tool_search_code"]);
+    expect(result.sdkTools.map((tool) => tool.name)).toEqual(["tool_search_code"]);
+  });
+
+  it("filters the hidden tool_search catalog before compacting narrowed tools", async () => {
+    let catalogRef: { current?: { entries?: Array<{ name: string }> } } | undefined;
+    const createOpenClawCodingTools = vi.fn(async (opts: unknown) => {
+      catalogRef = (opts as { toolSearchCatalogRef?: typeof catalogRef }).toolSearchCatalogRef;
+      return [
+        makeTool({ name: "tool_search_code" }),
+        makeTool({ name: "read" }),
+        makeTool({ name: "edit" }),
+        makeTool({ name: "write" }),
+      ];
+    });
+
+    await createCopilotToolBridge({
+      agentId: "agent-1",
+      attemptParams: {
+        config: { tools: { toolSearch: true } },
+        runId: "run-tool-search",
+        sessionKey: "agent:main:main",
+        toolsAllow: ["read"],
+      } as never,
+      createOpenClawCodingTools,
+      modelId: "gpt-4o",
+      modelProvider: "github-copilot",
+      sessionId: "session-1",
+    });
+
+    expect(catalogRef?.current?.entries?.map((entry) => entry.name)).toEqual(["read"]);
+  });
+
+  it("compacts the Copilot tool surface behind code-mode exec/wait when enabled", async () => {
+    const createOpenClawCodingTools = vi.fn(async () => [
+      makeTool({ name: "fake_hidden" }),
+      makeTool({ name: "read" }),
+    ]);
+
+    const result = await createCopilotToolBridge({
+      agentId: "agent-1",
+      attemptParams: {
+        config: { tools: { codeMode: true } },
+        runId: "run-code-mode",
+        sessionKey: "agent:main:main",
+      } as never,
+      createOpenClawCodingTools,
+      modelId: "gpt-4o",
+      modelProvider: "github-copilot",
+      sessionId: "session-1",
+    });
+
+    expect(createOpenClawCodingTools).toHaveBeenCalledWith(
+      expect.objectContaining({
+        includeToolSearchControls: false,
+        toolSearchCatalogRef: expect.any(Object),
+        toolSearchCatalogExecutor: expect.any(Function),
+      }),
+    );
+    expect(result.sourceTools.map((tool) => tool.name)).toEqual(["exec", "wait"]);
+    expect(result.sdkTools.map((tool) => tool.name)).toEqual(["exec", "wait"]);
+  });
+
+  it("keeps code-mode controls visible when a narrow allowlist is active", async () => {
+    const createOpenClawCodingTools = vi.fn(async () => [
+      makeTool({ name: "fake_hidden" }),
+      makeTool({ name: "read" }),
+    ]);
+
+    const result = await createCopilotToolBridge({
+      agentId: "agent-1",
+      attemptParams: {
+        config: { tools: { codeMode: true } },
+        runId: "run-code-mode",
+        sessionKey: "agent:main:main",
+        toolsAllow: ["read"],
+      } as never,
+      createOpenClawCodingTools,
+      modelId: "gpt-4o",
+      modelProvider: "github-copilot",
+      sessionId: "session-1",
+    });
+
+    expect(result.sourceTools.map((tool) => tool.name)).toEqual(["exec", "wait"]);
+    expect(result.sdkTools.map((tool) => tool.name)).toEqual(["exec", "wait"]);
+  });
+
+  it("filters the hidden code-mode catalog before compacting narrowed tools", async () => {
+    let catalogRef: { current?: { entries?: Array<{ name: string }> } } | undefined;
+    const createOpenClawCodingTools = vi.fn(async (opts: unknown) => {
+      catalogRef = (opts as { toolSearchCatalogRef?: typeof catalogRef }).toolSearchCatalogRef;
+      return [makeTool({ name: "read" }), makeTool({ name: "edit" }), makeTool({ name: "write" })];
+    });
+
+    await createCopilotToolBridge({
+      agentId: "agent-1",
+      attemptParams: {
+        config: { tools: { codeMode: true } },
+        runId: "run-code-mode",
+        sessionKey: "agent:main:main",
+        toolsAllow: ["read"],
+      } as never,
+      createOpenClawCodingTools,
+      modelId: "gpt-4o",
+      modelProvider: "github-copilot",
+      sessionId: "session-1",
+    });
+
+    expect(catalogRef?.current?.entries?.map((entry) => entry.name)).toEqual(["read"]);
+  });
+
   it("throws when createOpenClawCodingTools returns a non-array", async () => {
     await expect(
       createCopilotToolBridge({

extensions/copilot/src/tool-bridge.ts

@@ -17,10 +17,15 @@ import {
   resolveModelAuthMode,
   sanitizeToolResult,
 } from "openclaw/plugin-sdk/agent-harness-runtime";
+import { createAgentHarnessToolSurfaceRuntime } from "openclaw/plugin-sdk/agent-harness-tool-runtime";
 
 type CreateOpenClawCodingTools =
   (typeof import("openclaw/plugin-sdk/agent-harness"))["createOpenClawCodingTools"];
 type OpenClawCodingToolsOptions = NonNullable<Parameters<CreateOpenClawCodingTools>[0]>;
+type AgentHarnessToolSurfaceRuntime = ReturnType<typeof createAgentHarnessToolSurfaceRuntime>;
+type CatalogExecuteParams = Parameters<
+  NonNullable<AgentHarnessToolSurfaceRuntime["toolSearchCatalogExecutor"]>
+>[0];
 
 type AgentToolResultLike = {
   content?: unknown;
@@ -130,6 +135,7 @@ export interface CopilotToolBridgeInput {
 }
 
 export interface CopilotToolBridge {
+  cleanup?: () => void;
   sdkTools: SdkTool[];
   sourceTools: AnyAgentTool[];
 }
@@ -178,7 +184,31 @@ export async function createCopilotToolBridge(
     input.createOpenClawCodingTools ??
     (await import("openclaw/plugin-sdk/agent-harness")).createOpenClawCodingTools;
 
-  const toolOptions = buildOpenClawCodingToolsOptions(input, effectiveToolPlan);
+  const toolSurfaceRuntime = createAgentHarnessToolSurfaceRuntime({
+    abortSignal: input.abortSignal,
+    agentId: input.agentId,
+    config: attemptParams.config,
+    disableTools: attemptParams.disableTools,
+    executeTool: (toolParams) => executeCatalogTool(input, toolParams),
+    forceMessageTool: shouldForceCopilotMessageTool(attemptParams),
+    isRawModelRun: isCopilotRawModelRun(attemptParams),
+    modelToolsEnabled: true,
+    prompt: attemptParams.prompt,
+    runId: attemptParams.runId,
+    runtimeToolAllowlist: effectiveToolPlan.runtimeToolAllowlist,
+    sessionId: input.sessionId,
+    sessionKey: attemptParams.sandboxSessionKey ?? attemptParams.sessionKey ?? input.sessionKey,
+    sourceReplyDeliveryMode: attemptParams.sourceReplyDeliveryMode,
+    toolsAllow: attemptParams.toolsAllow,
+  });
+  const toolOptions = buildOpenClawCodingToolsOptions(
+    input,
+    {
+      ...effectiveToolPlan,
+      runtimeToolAllowlist: toolSurfaceRuntime.runtimeToolAllowlist,
+    },
+    toolSurfaceRuntime,
+  );
 
   let sourceTools: unknown;
   try {
@@ -196,13 +226,19 @@ export async function createCopilotToolBridge(
     );
   }
 
-  const plannedTools = filterCopilotToolsForConstructionPlan(
+  const allowedSourceTools = filterCopilotToolsForAllowlist(
     sourceTools as AnyAgentTool[],
+    toolSurfaceRuntime.runtimeToolAllowlist,
+  );
+  const compactedTools = toolSurfaceRuntime.compactTools(allowedSourceTools);
+  const plannedTools = filterCopilotToolsForConstructionPlan(
+    compactedTools.tools,
     effectiveToolPlan.codingToolConstructionPlan,
+    { preserveToolNames: toolSurfaceRuntime.runtimeToolAllowlist },
   );
   const filteredTools = filterCopilotToolsForAllowlist(
     plannedTools,
-    effectiveToolPlan.runtimeToolAllowlist,
+    toolSurfaceRuntime.runtimeToolAllowlist,
   );
 
   // Run duplicate detection after filtering so a duplicate in a
@@ -214,6 +250,7 @@ export async function createCopilotToolBridge(
   }
 
   return {
+    cleanup: toolSurfaceRuntime.cleanup,
     sdkTools: filteredTools.map((sourceTool) =>
       convertOpenClawToolToSdkTool(sourceTool, {
         abortSignal: input.abortSignal,
@@ -251,6 +288,7 @@ export async function createCopilotToolBridge(
 function buildOpenClawCodingToolsOptions(
   input: CopilotToolBridgeInput,
   toolPlan: ReturnType<typeof resolveEmbeddedAttemptToolConstructionPlan>,
+  toolSurfaceRuntime?: ReturnType<typeof createAgentHarnessToolSurfaceRuntime>,
 ): OpenClawCodingToolsOptions {
   const a = input.attemptParams ?? ({} as CopilotToolAttemptParams);
 
@@ -339,11 +377,14 @@ function buildOpenClawCodingToolsOptions(
     // `resolveSandboxContext`).
     sandbox,
     spawnWorkspaceDir,
-    config: a.config,
+    config: toolSurfaceRuntime?.config ?? a.config,
     abortSignal: input.abortSignal,
     modelProvider: input.modelProvider,
     modelId: input.modelId,
     includeCoreTools: toolPlan.includeCoreTools,
+    includeToolSearchControls: toolSurfaceRuntime?.includeToolSearchControls,
+    toolSearchCatalogRef: toolSurfaceRuntime?.toolSearchCatalogRef,
+    toolSearchCatalogExecutor: toolSurfaceRuntime?.toolSearchCatalogExecutor,
     runtimeToolAllowlist: toolPlan.runtimeToolAllowlist,
     toolConstructionPlan: toolPlan.codingToolConstructionPlan,
     modelCompat,
@@ -575,6 +616,63 @@ export function convertOpenClawToolToSdkTool(
   };
 }
 
+async function executeCatalogTool(
+  input: CopilotToolBridgeInput,
+  params: CatalogExecuteParams,
+): Promise<Awaited<ReturnType<AnyAgentTool["execute"]>>> {
+  const sourceTool = params.tool as AnyAgentTool;
+  const startedAt = Date.now();
+  let preparedArgs: unknown = params.input;
+  try {
+    preparedArgs = sourceTool.prepareArguments
+      ? sourceTool.prepareArguments(params.input)
+      : params.input;
+    const result = await sourceTool.execute(
+      params.toolCallId,
+      preparedArgs,
+      params.signal ?? input.abortSignal,
+      params.onUpdate,
+    );
+    const sanitizedResult = sanitizeToolResult(result);
+    const isError = isToolResultError(sanitizedResult);
+    input.attemptParams?.onAgentToolResult?.({
+      toolName: params.toolName,
+      result: sanitizedResult,
+      isError,
+    });
+    await input.onToolCompleted?.({
+      toolName: params.toolName,
+      toolCallId: params.toolCallId,
+      args: toToolStartArgs(preparedArgs),
+      result: sanitizedResult,
+      ...(isError
+        ? { error: extractToolErrorMessage(sanitizedResult) ?? "tool returned an error" }
+        : {}),
+      startedAt,
+    });
+    return result;
+  } catch (error: unknown) {
+    const message = toError(error).message;
+    const failure = sanitizeToolResult({
+      content: [{ type: "text", text: message }],
+      details: { status: "failed", error: message },
+    });
+    input.attemptParams?.onAgentToolResult?.({
+      toolName: params.toolName,
+      result: failure,
+      isError: true,
+    });
+    await input.onToolCompleted?.({
+      toolName: params.toolName,
+      toolCallId: params.toolCallId,
+      args: toToolStartArgs(preparedArgs),
+      error: message,
+      startedAt,
+    });
+    throw error;
+  }
+}
+
 function toToolStartArgs(args: unknown): Record<string, unknown> {
   return args && typeof args === "object" && !Array.isArray(args)
     ? (args as Record<string, unknown>)
@@ -712,11 +810,16 @@ function filterCopilotToolsForAllowlist<T extends { name: string }>(
 function filterCopilotToolsForConstructionPlan<T extends { name: string }>(
   tools: T[],
   plan: ReturnType<typeof resolveEmbeddedAttemptToolConstructionPlan>["codingToolConstructionPlan"],
+  options: { preserveToolNames?: readonly string[] } = {},
 ): T[] {
   if (plan.includeBaseCodingTools && plan.includeShellTools) {
     return tools;
   }
+  const preserveToolNames = new Set(options.preserveToolNames);
   return tools.filter((tool) => {
+    if (preserveToolNames.has(tool.name)) {
+      return true;
+    }
     if (!plan.includeBaseCodingTools && BASE_COPILOT_CODING_TOOL_NAMES.has(tool.name)) {
       return false;
     }

extensions/copilot/src/user-input-bridge.test.ts

@@ -0,0 +1,121 @@
+// Copilot tests cover SDK ask_user bridge behavior.
+import type { EmbeddedRunAttemptParams } from "openclaw/plugin-sdk/agent-harness-runtime";
+import { describe, expect, it, vi } from "vitest";
+import { createCopilotUserInputBridge } from "./user-input-bridge.js";
+
+function createParams(): EmbeddedRunAttemptParams {
+  return {
+    sessionId: "session-1",
+    sessionKey: "agent:main:session-1",
+    onBlockReply: vi.fn(),
+  } as unknown as EmbeddedRunAttemptParams;
+}
+
+function expectFirstBlockReplyText(params: EmbeddedRunAttemptParams): string {
+  const onBlockReply = params.onBlockReply;
+  if (!onBlockReply) {
+    throw new Error("Expected onBlockReply callback");
+  }
+  const payload = vi.mocked(onBlockReply).mock.calls[0]?.[0];
+  if (typeof payload?.text !== "string") {
+    throw new Error("Expected first block reply text");
+  }
+  return payload.text;
+}
+
+describe("Copilot user input bridge", () => {
+  it("prompts through OpenClaw and resolves the SDK request from the next queued message", async () => {
+    const params = createParams();
+    const bridge = createCopilotUserInputBridge({ paramsForRun: params });
+
+    const response = bridge.onUserInputRequest(
+      {
+        question: "Pick a mode",
+        choices: ["Fast", "Deep"],
+        allowFreeform: false,
+      },
+      { sessionId: "sdk-session-1" },
+    );
+
+    await vi.waitFor(() => expect(params.onBlockReply).toHaveBeenCalledTimes(1));
+    expect(expectFirstBlockReplyText(params)).toContain("Pick a mode");
+    expect(bridge.handleQueuedMessage("2")).toBe(true);
+
+    await expect(response).resolves.toEqual({ answer: "Deep", wasFreeform: false });
+  });
+
+  it("returns free-form answers when Copilot allows them", async () => {
+    const params = createParams();
+    const bridge = createCopilotUserInputBridge({ paramsForRun: params });
+
+    const response = bridge.onUserInputRequest(
+      {
+        question: "Which branch?",
+        allowFreeform: true,
+      },
+      { sessionId: "sdk-session-1" },
+    );
+
+    await vi.waitFor(() => expect(params.onBlockReply).toHaveBeenCalledTimes(1));
+    expect(bridge.handleQueuedMessage("fix/harness-parity")).toBe(true);
+
+    await expect(response).resolves.toEqual({
+      answer: "fix/harness-parity",
+      wasFreeform: true,
+    });
+  });
+
+  it("escapes SDK-controlled prompt text before channel delivery", async () => {
+    const params = createParams();
+    const bridge = createCopilotUserInputBridge({ paramsForRun: params });
+
+    void bridge.onUserInputRequest(
+      {
+        question: "Pick [trusted](https://evil) <@U123> @here\u202e",
+        choices: ["One @everyone", "Two `code`"],
+        allowFreeform: false,
+      },
+      { sessionId: "sdk-session-1" },
+    );
+
+    await vi.waitFor(() => expect(params.onBlockReply).toHaveBeenCalledTimes(1));
+    const text = expectFirstBlockReplyText(params);
+    expect(text).not.toContain("@here");
+    expect(text).not.toContain("@everyone");
+    expect(text).not.toContain("<@U123>");
+    expect(text).not.toContain("[trusted](https://evil)");
+    expect(text).not.toContain("`code`");
+    expect(text).toContain("\uff20here");
+    expect(text).toContain("\uff3btrusted\uff3d");
+  });
+
+  it("rejects queued messages when no ask_user request is pending", () => {
+    const bridge = createCopilotUserInputBridge({ paramsForRun: createParams() });
+
+    expect(bridge.handleQueuedMessage("late")).toBe(false);
+  });
+
+  it("resolves pending requests with an empty answer when aborted", async () => {
+    const params = createParams();
+    const controller = new AbortController();
+    const bridge = createCopilotUserInputBridge({
+      paramsForRun: params,
+      signal: controller.signal,
+    });
+
+    const response = bridge.onUserInputRequest(
+      {
+        question: "Continue?",
+        choices: ["Yes", "No"],
+        allowFreeform: false,
+      },
+      { sessionId: "sdk-session-1" },
+    );
+
+    await vi.waitFor(() => expect(params.onBlockReply).toHaveBeenCalledTimes(1));
+    controller.abort();
+
+    await expect(response).resolves.toEqual({ answer: "", wasFreeform: true });
+    expect(bridge.handleQueuedMessage("1")).toBe(false);
+  });
+});

extensions/copilot/src/user-input-bridge.ts

@@ -0,0 +1,161 @@
+import type { SessionConfig } from "@github/copilot-sdk";
+import {
+  buildAgentHarnessUserInputAnswers,
+  deliverAgentHarnessUserInputPrompt,
+  embeddedAgentLog,
+  type AgentHarnessUserInputQuestion,
+  type EmbeddedRunAttemptParams,
+} from "openclaw/plugin-sdk/agent-harness-runtime";
+
+type PendingCopilotUserInput = {
+  question: AgentHarnessUserInputQuestion;
+  resolve: (value: CopilotUserInputResponse) => void;
+  cleanup: () => void;
+};
+
+type CopilotUserInputHandler = NonNullable<SessionConfig["onUserInputRequest"]>;
+type CopilotUserInputRequest = Parameters<CopilotUserInputHandler>[0];
+type CopilotUserInputResponse = Awaited<ReturnType<CopilotUserInputHandler>>;
+
+type CopilotUserInputBridge = {
+  onUserInputRequest: CopilotUserInputHandler;
+  handleQueuedMessage: (text: string) => boolean;
+  cancelPending: () => void;
+};
+
+const COPILOT_USER_INPUT_QUESTION_ID = "answer";
+
+export function createCopilotUserInputBridge(params: {
+  paramsForRun: EmbeddedRunAttemptParams;
+  signal?: AbortSignal;
+}): CopilotUserInputBridge {
+  let pending: PendingCopilotUserInput | undefined;
+
+  const resolvePending = (value: CopilotUserInputResponse) => {
+    const current = pending;
+    if (!current) {
+      return;
+    }
+    pending = undefined;
+    current.cleanup();
+    current.resolve(value);
+  };
+
+  return {
+    onUserInputRequest(request) {
+      const question = toQuestion(request);
+      resolvePending(emptyCopilotUserInputResponse());
+      return new Promise<CopilotUserInputResponse>((resolve) => {
+        const abortListener = () => resolvePending(emptyCopilotUserInputResponse());
+        const cleanup = () => params.signal?.removeEventListener("abort", abortListener);
+        pending = { question, resolve, cleanup };
+        params.signal?.addEventListener("abort", abortListener, { once: true });
+        if (params.signal?.aborted) {
+          resolvePending(emptyCopilotUserInputResponse());
+          return;
+        }
+        void deliverAgentHarnessUserInputPrompt(params.paramsForRun, [question], {
+          intro: "Copilot needs input:",
+          formatText: formatCopilotDisplayText,
+        }).catch((error: unknown) => {
+          embeddedAgentLog.warn("failed to deliver copilot user input prompt", { error });
+        });
+      });
+    },
+    handleQueuedMessage(text) {
+      const current = pending;
+      if (!current) {
+        return false;
+      }
+      resolvePending(buildCopilotUserInputResponse(current.question, text));
+      return true;
+    },
+    cancelPending() {
+      resolvePending(emptyCopilotUserInputResponse());
+    },
+  };
+}
+
+function toQuestion(request: CopilotUserInputRequest): AgentHarnessUserInputQuestion {
+  return {
+    id: COPILOT_USER_INPUT_QUESTION_ID,
+    header: "Copilot needs input",
+    question: request.question,
+    isOther: request.allowFreeform !== false,
+    isSecret: false,
+    options:
+      request.choices && request.choices.length > 0
+        ? request.choices.map((choice: string) => ({ label: choice }))
+        : null,
+  };
+}
+
+function buildCopilotUserInputResponse(
+  question: AgentHarnessUserInputQuestion,
+  inputText: string,
+): CopilotUserInputResponse {
+  const rawAnswers = buildAgentHarnessUserInputAnswers([question], inputText);
+  const selected = rawAnswers.answers[COPILOT_USER_INPUT_QUESTION_ID]?.answers[0] ?? "";
+  return {
+    answer: selected,
+    wasFreeform: !isChoiceAnswer(question, selected),
+  };
+}
+
+function emptyCopilotUserInputResponse(): CopilotUserInputResponse {
+  return { answer: "", wasFreeform: true };
+}
+
+function isChoiceAnswer(question: AgentHarnessUserInputQuestion, answer: string): boolean {
+  return Boolean(
+    answer &&
+    question.options?.some((option) => option.label.toLowerCase() === answer.toLowerCase()),
+  );
+}
+
+function formatCopilotDisplayText(value: string): string {
+  const safe = sanitizeCopilotDisplayText(value).trim();
+  return escapeCopilotChatText(safe || "<unknown>");
+}
+
+function sanitizeCopilotDisplayText(value: string): string {
+  let safe = "";
+  for (const character of value) {
+    const codePoint = character.codePointAt(0);
+    safe += codePoint != null && isUnsafeDisplayCodePoint(codePoint) ? "?" : character;
+  }
+  return safe;
+}
+
+function escapeCopilotChatText(value: string): string {
+  return value
+    .replaceAll("&", "&amp;")
+    .replaceAll("<", "&lt;")
+    .replaceAll(">", "&gt;")
+    .replaceAll("@", "\uff20")
+    .replaceAll("`", "\uff40")
+    .replaceAll("[", "\uff3b")
+    .replaceAll("]", "\uff3d")
+    .replaceAll("(", "\uff08")
+    .replaceAll(")", "\uff09")
+    .replaceAll("*", "\u2217")
+    .replaceAll("_", "\uff3f")
+    .replaceAll("~", "\uff5e")
+    .replaceAll("|", "\uff5c");
+}
+
+function isUnsafeDisplayCodePoint(codePoint: number): boolean {
+  return (
+    codePoint <= 0x001f ||
+    (codePoint >= 0x007f && codePoint <= 0x009f) ||
+    codePoint === 0x00ad ||
+    codePoint === 0x061c ||
+    codePoint === 0x180e ||
+    (codePoint >= 0x200b && codePoint <= 0x200f) ||
+    (codePoint >= 0x202a && codePoint <= 0x202e) ||
+    (codePoint >= 0x2060 && codePoint <= 0x206f) ||
+    codePoint === 0xfeff ||
+    (codePoint >= 0xfff9 && codePoint <= 0xfffb) ||
+    (codePoint >= 0xe0000 && codePoint <= 0xe007f)
+  );
+}

extensions/deepinfra/index.test.ts

@@ -49,6 +49,14 @@ function makeAgentModelEntry(id = "profile/live-model") {
   };
 }
 
+function jsonResponse(payload: unknown, init: ResponseInit = {}): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "Content-Type": "application/json" },
+    ...init,
+  });
+}
+
 async function withLiveDiscoveryTestEnv(
   mockFetch: ReturnType<typeof vi.fn>,
   runAssertions: () => Promise<void>,
@@ -122,10 +130,9 @@ describe("deepinfra augmentModelCatalog", () => {
 
   it("uses config-backed API keys to enable live model catalog augmentation", async () => {
     resetDeepInfraModelCacheForTest();
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () => Promise.resolve({ data: [makeAgentModelEntry("config/live-model")] }),
-    });
+    const mockFetch = vi
+      .fn()
+      .mockResolvedValue(jsonResponse({ data: [makeAgentModelEntry("config/live-model")] }));
     const provider = await registerSingleProviderPlugin(deepinfraPlugin);
 
     await withLiveDiscoveryTestEnv(mockFetch, async () => {
@@ -151,10 +158,9 @@ describe("deepinfra augmentModelCatalog", () => {
 
   it("still runs live discovery when ctx.entries includes custom DeepInfra rows", async () => {
     resetDeepInfraModelCacheForTest();
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () => Promise.resolve({ data: [makeAgentModelEntry("custom/live-model")] }),
-    });
+    const mockFetch = vi
+      .fn()
+      .mockResolvedValue(jsonResponse({ data: [makeAgentModelEntry("custom/live-model")] }));
     const provider = await registerSingleProviderPlugin(deepinfraPlugin);
 
     const seededDeepInfraCount = DEEPINFRA_MODEL_CATALOG.length + 5;
@@ -230,10 +236,7 @@ describe("deepinfra capability registration", () => {
 
   it("uses profile-resolved API keys for live text catalog discovery", async () => {
     resetDeepInfraModelCacheForTest();
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () => Promise.resolve({ data: [makeAgentModelEntry()] }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(jsonResponse({ data: [makeAgentModelEntry()] }));
     const captured = createCapturedPluginRegistration();
     deepinfraPlugin.register(captured.api);
     const provider = captured.providers[0];

extensions/deepinfra/provider-models.test.ts

@@ -48,6 +48,14 @@ function makeAgentModelEntry(overrides: Record<string, unknown> = {}) {
   };
 }
 
+function jsonResponse(payload: unknown, init: ResponseInit = {}): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "Content-Type": "application/json" },
+    ...init,
+  });
+}
+
 function expectedStaticChatCatalog() {
   return DEEPINFRA_MODEL_CATALOG.map((model) => {
     const compat = Object.assign({}, model.compat, {
@@ -195,10 +203,7 @@ describe("discoverDeepInfraModels (chat-only shim)", () => {
   });
 
   it("fetches the openclaw-projection endpoint and parses chat-surface entries when an API key is configured", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () => Promise.resolve({ data: [makeAgentModelEntry()] }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(jsonResponse({ data: [makeAgentModelEntry()] }));
 
     await withFetchPathTest(mockFetch, { DEEPINFRA_API_KEY: "sk-test" }, async () => {
       const models = await discoverDeepInfraModels();
@@ -228,21 +233,19 @@ describe("discoverDeepInfraModels (chat-only shim)", () => {
   });
 
   it("skips entries with no metadata or no surface tag, and deduplicates ids", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            { id: "BAAI/bge-m3", object: "model", metadata: null },
-            makeAgentModelEntry({
-              id: "untagged/model",
-              metadata: { context_length: 1, max_tokens: 1, pricing: {}, tags: [] },
-            }),
-            makeAgentModelEntry(),
-            makeAgentModelEntry(),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          { id: "BAAI/bge-m3", object: "model", metadata: null },
+          makeAgentModelEntry({
+            id: "untagged/model",
+            metadata: { context_length: 1, max_tokens: 1, pricing: {}, tags: [] },
+          }),
+          makeAgentModelEntry(),
+          makeAgentModelEntry(),
+        ],
+      }),
+    );
 
     await withFetchPathTest(mockFetch, { DEEPINFRA_API_KEY: "sk-test" }, async () => {
       const models = await discoverDeepInfraModels();
@@ -283,7 +286,7 @@ describe("discoverDeepInfraModels (chat-only shim)", () => {
   });
 
   it("falls back to the static catalog on non-2xx HTTP responses", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({ ok: false, status: 503 });
+    const mockFetch = vi.fn().mockResolvedValue(new Response("", { status: 503 }));
 
     await withFetchPathTest(mockFetch, { DEEPINFRA_API_KEY: "sk-test" }, async () => {
       const models = await discoverDeepInfraModels();
@@ -294,14 +297,10 @@ describe("discoverDeepInfraModels (chat-only shim)", () => {
   it("falls back without caching malformed successful model list payloads", async () => {
     const mockFetch = vi
       .fn()
-      .mockResolvedValueOnce({
-        ok: true,
-        json: () => Promise.resolve({ data: {} }),
-      })
-      .mockResolvedValueOnce({
-        ok: true,
-        json: () => Promise.resolve({ data: [makeAgentModelEntry({ id: "recovered/model" })] }),
-      });
+      .mockResolvedValueOnce(jsonResponse({ data: {} }))
+      .mockResolvedValueOnce(
+        jsonResponse({ data: [makeAgentModelEntry({ id: "recovered/model" })] }),
+      );
 
     await withFetchPathTest(mockFetch, { DEEPINFRA_API_KEY: "sk-test" }, async () => {
       expect((await discoverDeepInfraModels()).map((m) => m.id)).toEqual(
@@ -328,14 +327,8 @@ describe("discoverDeepInfraModels (chat-only shim)", () => {
   it("caches successful discovery responses only", async () => {
     const mockFetch = vi
       .fn()
-      .mockResolvedValueOnce({
-        ok: true,
-        json: () => Promise.resolve({ data: [makeAgentModelEntry({ id: "first/model" })] }),
-      })
-      .mockResolvedValueOnce({
-        ok: true,
-        json: () => Promise.resolve({ data: [makeAgentModelEntry({ id: "second/model" })] }),
-      });
+      .mockResolvedValueOnce(jsonResponse({ data: [makeAgentModelEntry({ id: "first/model" })] }))
+      .mockResolvedValueOnce(jsonResponse({ data: [makeAgentModelEntry({ id: "second/model" })] }));
 
     await withFetchPathTest(mockFetch, { DEEPINFRA_API_KEY: "sk-test" }, async () => {
       const expectedIds = expectedLiveChatCatalog([
@@ -359,14 +352,10 @@ describe("discoverDeepInfraModels (chat-only shim)", () => {
   it("does not cache successful responses that produce no live catalog rows", async () => {
     const mockFetch = vi
       .fn()
-      .mockResolvedValueOnce({
-        ok: true,
-        json: () => Promise.resolve({ data: [] }),
-      })
-      .mockResolvedValueOnce({
-        ok: true,
-        json: () => Promise.resolve({ data: [makeAgentModelEntry({ id: "recovered/model" })] }),
-      });
+      .mockResolvedValueOnce(jsonResponse({ data: [] }))
+      .mockResolvedValueOnce(
+        jsonResponse({ data: [makeAgentModelEntry({ id: "recovered/model" })] }),
+      );
 
     await withFetchPathTest(mockFetch, { DEEPINFRA_API_KEY: "sk-test" }, async () => {
       expect((await discoverDeepInfraModels()).map((m) => m.id)).toEqual(
@@ -393,67 +382,65 @@ describe("discoverDeepInfraModels (chat-only shim)", () => {
 
 describe("discoverDeepInfraSurfaces (per-surface bucketing)", () => {
   it("buckets dynamic entries by short-alias surface tag", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            makeAgentModelEntry({
-              id: "anthropic/claude-sonnet-4-6",
-              metadata: {
-                description: "claude sonnet 4.6",
-                context_length: 200000,
-                max_tokens: 8192,
-                pricing: { input_tokens: 3, output_tokens: 15 },
-                tags: ["chat", "vlm", "vision", "prompt_cache"],
-              },
-            }),
-            makeAgentModelEntry({
-              id: "BAAI/bge-m3",
-              metadata: {
-                description: "bge-m3",
-                pricing: { input_tokens: 0.01 },
-                tags: ["embed"],
-              },
-            }),
-            makeAgentModelEntry({
-              id: "black-forest-labs/FLUX-1-schnell",
-              metadata: {
-                description: "FLUX schnell",
-                pricing: { per_image_unit: 0.003 },
-                tags: ["image-gen"],
-                default_width: 1024,
-                default_height: 1024,
-                default_iterations: 4,
-              },
-            }),
-            makeAgentModelEntry({
-              id: "Wan-AI/Wan2.6-T2V",
-              metadata: {
-                description: "Wan T2V",
-                pricing: { output_seconds: 0.05 },
-                tags: ["video-gen"],
-              },
-            }),
-            makeAgentModelEntry({
-              id: "Qwen/Qwen3-TTS",
-              metadata: {
-                description: "Qwen3 TTS",
-                pricing: { input_characters: 0.65 },
-                tags: ["tts"],
-              },
-            }),
-            makeAgentModelEntry({
-              id: "openai/whisper-large-v3-turbo",
-              metadata: {
-                description: "whisper",
-                pricing: { input_seconds: 0.00004 },
-                tags: ["stt"],
-              },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          makeAgentModelEntry({
+            id: "anthropic/claude-sonnet-4-6",
+            metadata: {
+              description: "claude sonnet 4.6",
+              context_length: 200000,
+              max_tokens: 8192,
+              pricing: { input_tokens: 3, output_tokens: 15 },
+              tags: ["chat", "vlm", "vision", "prompt_cache"],
+            },
+          }),
+          makeAgentModelEntry({
+            id: "BAAI/bge-m3",
+            metadata: {
+              description: "bge-m3",
+              pricing: { input_tokens: 0.01 },
+              tags: ["embed"],
+            },
+          }),
+          makeAgentModelEntry({
+            id: "black-forest-labs/FLUX-1-schnell",
+            metadata: {
+              description: "FLUX schnell",
+              pricing: { per_image_unit: 0.003 },
+              tags: ["image-gen"],
+              default_width: 1024,
+              default_height: 1024,
+              default_iterations: 4,
+            },
+          }),
+          makeAgentModelEntry({
+            id: "Wan-AI/Wan2.6-T2V",
+            metadata: {
+              description: "Wan T2V",
+              pricing: { output_seconds: 0.05 },
+              tags: ["video-gen"],
+            },
+          }),
+          makeAgentModelEntry({
+            id: "Qwen/Qwen3-TTS",
+            metadata: {
+              description: "Qwen3 TTS",
+              pricing: { input_characters: 0.65 },
+              tags: ["tts"],
+            },
+          }),
+          makeAgentModelEntry({
+            id: "openai/whisper-large-v3-turbo",
+            metadata: {
+              description: "whisper",
+              pricing: { input_seconds: 0.00004 },
+              tags: ["stt"],
+            },
+          }),
+        ],
+      }),
+    );
 
     await withFetchPathTest(mockFetch, { DEEPINFRA_API_KEY: "sk-test" }, async () => {
       const catalog = await discoverDeepInfraSurfaces();
@@ -471,35 +458,33 @@ describe("discoverDeepInfraSurfaces (per-surface bucketing)", () => {
   });
 
   it("drops malformed live numeric metadata", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            makeAgentModelEntry({
-              id: "bad/chat",
-              metadata: {
-                description: "bad chat",
-                context_length: -1,
-                max_tokens: 1.5,
-                pricing: { input_tokens: 3, output_tokens: 15 },
-                tags: ["chat"],
-              },
-            }),
-            makeAgentModelEntry({
-              id: "bad/image",
-              metadata: {
-                description: "bad image",
-                pricing: { per_image_unit: 0.003 },
-                tags: ["image-gen"],
-                default_width: Number.POSITIVE_INFINITY,
-                default_height: 1024.5,
-                default_iterations: 0,
-              },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          makeAgentModelEntry({
+            id: "bad/chat",
+            metadata: {
+              description: "bad chat",
+              context_length: -1,
+              max_tokens: 1.5,
+              pricing: { input_tokens: 3, output_tokens: 15 },
+              tags: ["chat"],
+            },
+          }),
+          makeAgentModelEntry({
+            id: "bad/image",
+            metadata: {
+              description: "bad image",
+              pricing: { per_image_unit: 0.003 },
+              tags: ["image-gen"],
+              default_width: Number.POSITIVE_INFINITY,
+              default_height: 1024.5,
+              default_iterations: 0,
+            },
+          }),
+        ],
+      }),
+    );
 
     await withFetchPathTest(mockFetch, { DEEPINFRA_API_KEY: "sk-test" }, async () => {
       const catalog = await discoverDeepInfraSurfaces();

extensions/deepinfra/surface-model-catalogs.test.ts

@@ -49,6 +49,14 @@ const surfaceEntry = (id: string, surfaceTag: string, extra: Record<string, unkn
   },
 });
 
+function jsonResponse(payload: unknown, init: ResponseInit = {}): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "Content-Type": "application/json" },
+    ...init,
+  });
+}
+
 async function withLiveFetch(mockFetch: ReturnType<typeof vi.fn>, run: () => Promise<void>) {
   const env = { ...process.env };
   delete process.env.NODE_ENV;
@@ -86,19 +94,17 @@ describe("DeepInfra generation catalogs", () => {
 
 describe("listDeepInfraImageGenCatalog", () => {
   it("returns null when live discovery succeeds but the response has zero image-gen entries", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            surfaceEntry("anthropic/claude-sonnet-4-6", "chat", {
-              context_length: 200000,
-              max_tokens: 8192,
-              pricing: { input_tokens: 3, output_tokens: 15 },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          surfaceEntry("anthropic/claude-sonnet-4-6", "chat", {
+            context_length: 200000,
+            max_tokens: 8192,
+            pricing: { input_tokens: 3, output_tokens: 15 },
+          }),
+        ],
+      }),
+    );
 
     await withLiveFetch(mockFetch, async () => {
       const result = await listDeepInfraImageGenCatalog(withKeyCtx());
@@ -115,28 +121,26 @@ describe("listDeepInfraImageGenCatalog", () => {
   });
 
   it("projects discovered image-gen entries when a key is configured and discovery is live", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            surfaceEntry("black-forest-labs/FLUX-2-pro", "image-gen", {
-              pricing: { per_image_unit: 0.08 },
-              default_width: 1024,
-              default_height: 1024,
-              default_iterations: 28,
-            }),
-            surfaceEntry("ByteDance/Seedream-4", "image-gen", {
-              pricing: { per_image_unit: 0.03 },
-            }),
-            surfaceEntry("anthropic/claude-sonnet-4-6", "chat", {
-              context_length: 200000,
-              max_tokens: 8192,
-              pricing: { input_tokens: 3, output_tokens: 15 },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          surfaceEntry("black-forest-labs/FLUX-2-pro", "image-gen", {
+            pricing: { per_image_unit: 0.08 },
+            default_width: 1024,
+            default_height: 1024,
+            default_iterations: 28,
+          }),
+          surfaceEntry("ByteDance/Seedream-4", "image-gen", {
+            pricing: { per_image_unit: 0.03 },
+          }),
+          surfaceEntry("anthropic/claude-sonnet-4-6", "chat", {
+            context_length: 200000,
+            max_tokens: 8192,
+            pricing: { input_tokens: 3, output_tokens: 15 },
+          }),
+        ],
+      }),
+    );
 
     await withLiveFetch(mockFetch, async () => {
       const result = await listDeepInfraImageGenCatalog(withKeyCtx());
@@ -161,22 +165,20 @@ describe("listDeepInfraVideoGenCatalog", () => {
     // produces zero video-gen entries. We must return null so the registered
     // provider's static fallback list is consulted instead of an empty
     // "live" answer.
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            surfaceEntry("anthropic/claude-sonnet-4-6", "chat", {
-              context_length: 200000,
-              max_tokens: 8192,
-              pricing: { input_tokens: 3, output_tokens: 15 },
-            }),
-            surfaceEntry("black-forest-labs/FLUX-2-pro", "image-gen", {
-              pricing: { per_image_unit: 0.08 },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          surfaceEntry("anthropic/claude-sonnet-4-6", "chat", {
+            context_length: 200000,
+            max_tokens: 8192,
+            pricing: { input_tokens: 3, output_tokens: 15 },
+          }),
+          surfaceEntry("black-forest-labs/FLUX-2-pro", "image-gen", {
+            pricing: { per_image_unit: 0.08 },
+          }),
+        ],
+      }),
+    );
 
     await withLiveFetch(mockFetch, async () => {
       const result = await listDeepInfraVideoGenCatalog(withKeyCtx());
@@ -185,20 +187,18 @@ describe("listDeepInfraVideoGenCatalog", () => {
   });
 
   it("projects discovered video-gen entries with capability shape", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            surfaceEntry("Wan-AI/Wan2.6-T2V", "video-gen", {
-              pricing: { output_seconds: 0.05 },
-            }),
-            surfaceEntry("ByteDance/Seedance-2.0", "video-gen", {
-              pricing: { output_seconds: 0.08 },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          surfaceEntry("Wan-AI/Wan2.6-T2V", "video-gen", {
+            pricing: { output_seconds: 0.05 },
+          }),
+          surfaceEntry("ByteDance/Seedance-2.0", "video-gen", {
+            pricing: { output_seconds: 0.08 },
+          }),
+        ],
+      }),
+    );
 
     await withLiveFetch(mockFetch, async () => {
       const result = await listDeepInfraVideoGenCatalog(withKeyCtx());
@@ -214,17 +214,15 @@ describe("listDeepInfraVideoGenCatalog", () => {
 
 describe("resolveDeepInfraVideoModelCapabilities", () => {
   it("returns capabilities for a discovered video-gen model", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            surfaceEntry("Wan-AI/Wan2.6-T2V", "video-gen", {
-              pricing: { output_seconds: 0.05 },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          surfaceEntry("Wan-AI/Wan2.6-T2V", "video-gen", {
+            pricing: { output_seconds: 0.05 },
+          }),
+        ],
+      }),
+    );
 
     await withLiveFetch(mockFetch, async () => {
       const caps = await resolveDeepInfraVideoModelCapabilities({
@@ -236,17 +234,15 @@ describe("resolveDeepInfraVideoModelCapabilities", () => {
   });
 
   it("strips the deepinfra/ prefix when matching", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            surfaceEntry("Wan-AI/Wan2.6-T2V", "video-gen", {
-              pricing: { output_seconds: 0.05 },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          surfaceEntry("Wan-AI/Wan2.6-T2V", "video-gen", {
+            pricing: { output_seconds: 0.05 },
+          }),
+        ],
+      }),
+    );
 
     await withLiveFetch(mockFetch, async () => {
       const caps = await resolveDeepInfraVideoModelCapabilities({
@@ -257,17 +253,15 @@ describe("resolveDeepInfraVideoModelCapabilities", () => {
   });
 
   it("returns undefined for an unknown model", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            surfaceEntry("Wan-AI/Wan2.6-T2V", "video-gen", {
-              pricing: { output_seconds: 0.05 },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          surfaceEntry("Wan-AI/Wan2.6-T2V", "video-gen", {
+            pricing: { output_seconds: 0.05 },
+          }),
+        ],
+      }),
+    );
 
     await withLiveFetch(mockFetch, async () => {
       const caps = await resolveDeepInfraVideoModelCapabilities({

extensions/github-copilot/models.test.ts

@@ -464,11 +464,7 @@ describe("fetchCopilotModelCatalog", () => {
   };
 
   it("maps Copilot /models entries to ModelDefinitionConfig with real context windows", async () => {
-    const fetchImpl = vi.fn().mockResolvedValue({
-      ok: true,
-      status: 200,
-      json: async () => sampleApiResponse,
-    });
+    const fetchImpl = vi.fn().mockResolvedValue(makeResponse(200, sampleApiResponse));
 
     const out = await fetchCopilotModelCatalog({
       copilotApiToken: "tid=test",
@@ -539,11 +535,7 @@ describe("fetchCopilotModelCatalog", () => {
   });
 
   it("strips trailing slash from baseUrl when building the /models URL", async () => {
-    const fetchImpl = vi.fn().mockResolvedValue({
-      ok: true,
-      status: 200,
-      json: async () => ({ data: [] }),
-    });
+    const fetchImpl = vi.fn().mockResolvedValue(makeResponse(200, { data: [] }));
 
     await fetchCopilotModelCatalog({
       copilotApiToken: "tid=test",
@@ -555,10 +547,8 @@ describe("fetchCopilotModelCatalog", () => {
   });
 
   it("dedupes by id when API returns duplicates", async () => {
-    const fetchImpl = vi.fn().mockResolvedValue({
-      ok: true,
-      status: 200,
-      json: async () => ({
+    const fetchImpl = vi.fn().mockResolvedValue(
+      makeResponse(200, {
         data: [
           {
             id: "gpt-5.5",
@@ -580,7 +570,7 @@ describe("fetchCopilotModelCatalog", () => {
           },
         ],
       }),
-    });
+    );
 
     const out = await fetchCopilotModelCatalog({
       copilotApiToken: "tid=test",
@@ -593,10 +583,8 @@ describe("fetchCopilotModelCatalog", () => {
   });
 
   it("falls back from malformed live token limits", async () => {
-    const fetchImpl = vi.fn().mockResolvedValue({
-      ok: true,
-      status: 200,
-      json: async () => ({
+    const fetchImpl = vi.fn().mockResolvedValue(
+      makeResponse(200, {
         data: [
           {
             id: "gpt-bad-window",
@@ -624,7 +612,7 @@ describe("fetchCopilotModelCatalog", () => {
           },
         ],
       }),
-    });
+    );
 
     const out = await fetchCopilotModelCatalog({
       copilotApiToken: "tid=test",
@@ -646,11 +634,7 @@ describe("fetchCopilotModelCatalog", () => {
   });
 
   it("throws on non-2xx HTTP responses so the caller can fall back to the static catalog", async () => {
-    const fetchImpl = vi.fn().mockResolvedValue({
-      ok: false,
-      status: 401,
-      json: async () => ({}),
-    });
+    const fetchImpl = vi.fn().mockResolvedValue(makeResponse(401, {}));
 
     await expect(
       fetchCopilotModelCatalog({
@@ -663,11 +647,7 @@ describe("fetchCopilotModelCatalog", () => {
 
   it("throws provider-owned errors for malformed successful /models payloads", async () => {
     for (const payload of [[], { data: {} }, { data: [null] }]) {
-      const fetchImpl = vi.fn().mockResolvedValue({
-        ok: true,
-        status: 200,
-        json: async () => payload,
-      });
+      const fetchImpl = vi.fn().mockResolvedValue(makeResponse(200, payload));
 
       await expect(
         fetchCopilotModelCatalog({

extensions/kilocode/provider-models.test.ts

@@ -14,16 +14,7 @@ vi.mock("openclaw/plugin-sdk/ssrf-runtime", () => ({
 
 import { discoverKilocodeModels, KILOCODE_MODELS_URL } from "./provider-models.js";
 
-type MockKilocodeFetchResponse = {
-  ok: boolean;
-  status?: number;
-  json?: () => Promise<unknown>;
-};
-
-type MockKilocodeFetch = ((
-  url: string,
-  init?: RequestInit,
-) => Promise<MockKilocodeFetchResponse>) & {
+type MockKilocodeFetch = ((url: string, init?: RequestInit) => Promise<Response>) & {
   mock: { calls: unknown[][] };
 };
 
@@ -115,6 +106,14 @@ function makeAutoModel(overrides: Record<string, unknown> = {}) {
   });
 }
 
+function jsonResponse(payload: unknown, init: ResponseInit = {}): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "Content-Type": "application/json" },
+    ...init,
+  });
+}
+
 async function withFetchPathTest(mockFetch: MockKilocodeFetch, runAssertions: () => Promise<void>) {
   const release = vi.fn(async () => {});
   vi.stubEnv("NODE_ENV", "");
@@ -165,13 +164,11 @@ describe("discoverKilocodeModels", () => {
 
 describe("discoverKilocodeModels (fetch path)", () => {
   it("parses gateway models with correct pricing conversion", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [makeAutoModel(), makeGatewayModel()],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [makeAutoModel(), makeGatewayModel()],
+      }),
+    );
     await withFetchPathTest(mockFetch, async () => {
       const models = await discoverKilocodeModels();
 
@@ -217,10 +214,7 @@ describe("discoverKilocodeModels (fetch path)", () => {
   });
 
   it("falls back to static catalog on HTTP error", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: false,
-      status: 500,
-    });
+    const mockFetch = vi.fn().mockResolvedValue(new Response("", { status: 500 }));
     await withFetchPathTest(mockFetch, async () => {
       const models = await discoverKilocodeModels();
       expect(models).toStrictEqual(EXPECTED_STATIC_KILOCODE_MODELS);
@@ -229,10 +223,7 @@ describe("discoverKilocodeModels (fetch path)", () => {
 
   it("falls back to static catalog for malformed successful model list payloads", async () => {
     for (const payload of [[], { data: {} }, { data: [null] }]) {
-      const mockFetch = vi.fn().mockResolvedValue({
-        ok: true,
-        json: () => Promise.resolve(payload),
-      });
+      const mockFetch = vi.fn().mockResolvedValue(jsonResponse(payload));
       await withFetchPathTest(mockFetch, async () => {
         const models = await discoverKilocodeModels();
         expect(models).toStrictEqual(EXPECTED_STATIC_KILOCODE_MODELS);
@@ -241,24 +232,22 @@ describe("discoverKilocodeModels (fetch path)", () => {
   });
 
   it("falls back from malformed live token metadata", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [
-            makeGatewayModel({
-              id: "some/bad-window",
-              context_length: -1,
-              top_provider: { max_completion_tokens: 8192.5 },
-            }),
-            makeGatewayModel({
-              id: "some/bad-output",
-              context_length: Number.POSITIVE_INFINITY,
-              top_provider: { max_completion_tokens: 0 },
-            }),
-          ],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [
+          makeGatewayModel({
+            id: "some/bad-window",
+            context_length: -1,
+            top_provider: { max_completion_tokens: 8192.5 },
+          }),
+          makeGatewayModel({
+            id: "some/bad-output",
+            context_length: Number.POSITIVE_INFINITY,
+            top_provider: { max_completion_tokens: 0 },
+          }),
+        ],
+      }),
+    );
 
     await withFetchPathTest(mockFetch, async () => {
       const models = await discoverKilocodeModels();
@@ -275,13 +264,11 @@ describe("discoverKilocodeModels (fetch path)", () => {
   });
 
   it("ensures kilo/auto is present even when API doesn't return it", async () => {
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [makeGatewayModel()],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [makeGatewayModel()],
+      }),
+    );
     await withFetchPathTest(mockFetch, async () => {
       const models = await discoverKilocodeModels();
       expect(requireModelById(models, "kilo/auto").id).toBe("kilo/auto");
@@ -301,10 +288,7 @@ describe("discoverKilocodeModels (fetch path)", () => {
       supported_parameters: ["max_tokens", "temperature"],
     });
 
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () => Promise.resolve({ data: [textOnlyModel] }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(jsonResponse({ data: [textOnlyModel] }));
     await withFetchPathTest(mockFetch, async () => {
       const models = await discoverKilocodeModels();
       const textModel = requireModelById(models, "some/text-model");
@@ -319,13 +303,11 @@ describe("discoverKilocodeModels (fetch path)", () => {
       pricing: undefined,
     });
 
-    const mockFetch = vi.fn().mockResolvedValue({
-      ok: true,
-      json: () =>
-        Promise.resolve({
-          data: [malformedAutoModel, makeAutoModel(), makeGatewayModel()],
-        }),
-    });
+    const mockFetch = vi.fn().mockResolvedValue(
+      jsonResponse({
+        data: [malformedAutoModel, makeAutoModel(), makeGatewayModel()],
+      }),
+    );
     await withFetchPathTest(mockFetch, async () => {
       const models = await discoverKilocodeModels();
       const auto = requireModelById(models, "kilo/auto");

extensions/lmstudio/src/models.test.ts

@@ -31,6 +31,21 @@ vi.mock("openclaw/plugin-sdk/ssrf-runtime", async (importOriginal) => {
   };
 });
 
+function jsonResponse(payload: unknown, init?: ResponseInit): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "content-type": "application/json" },
+    ...init,
+  });
+}
+
+function malformedJsonResponse(): Response {
+  return new Response("{ nope", {
+    status: 200,
+    headers: { "content-type": "application/json" },
+  });
+}
+
 afterAll(() => {
   vi.doUnmock("openclaw/plugin-sdk/ssrf-runtime");
   vi.resetModules();
@@ -72,33 +87,26 @@ describe("lmstudio-models", () => {
     loadedContextLength?: number;
     maxContextLength?: number;
   }) =>
-    vi.fn(async (url: string | URL, init?: RequestInit) => {
+    vi.fn(async (url: string | URL, _init?: RequestInit) => {
       const key = params?.key ?? "qwen3-8b-instruct";
       if (String(url).endsWith("/api/v1/models")) {
-        return {
-          ok: true,
-          json: async () => ({
-            models: [
-              {
-                type: "llm",
-                key,
-                max_context_length: params?.maxContextLength,
-                variants: params?.variants,
-                selected_variant: params?.selectedVariant,
-                loaded_instances: params?.loadedContextLength
-                  ? [{ id: "inst-1", config: { context_length: params.loadedContextLength } }]
-                  : [],
-              },
-            ],
-          }),
-        };
+        return jsonResponse({
+          models: [
+            {
+              type: "llm",
+              key,
+              max_context_length: params?.maxContextLength,
+              variants: params?.variants,
+              selected_variant: params?.selectedVariant,
+              loaded_instances: params?.loadedContextLength
+                ? [{ id: "inst-1", config: { context_length: params.loadedContextLength } }]
+                : [],
+            },
+          ],
+        });
       }
       if (String(url).endsWith("/api/v1/models/load")) {
-        return {
-          ok: true,
-          json: async () => ({ status: "loaded" }),
-          requestInit: init,
-        };
+        return jsonResponse({ status: "loaded" });
       }
       throw new Error(`Unexpected fetch URL: ${String(url)}`);
     });
@@ -296,9 +304,8 @@ describe("lmstudio-models", () => {
   });
 
   it("discovers llm models and maps metadata", async () => {
-    const fetchMock = vi.fn(async (_url: string | URL, _init?: RequestInit) => ({
-      ok: true,
-      json: async () => ({
+    const fetchMock = vi.fn(async (_url: string | URL, _init?: RequestInit) =>
+      jsonResponse({
         models: [
           {
             type: "llm",
@@ -330,7 +337,7 @@ describe("lmstudio-models", () => {
           },
         ],
       }),
-    }));
+    );
 
     const models = await discoverLmstudioModels({
       baseUrl: "http://localhost:1234/v1",
@@ -386,13 +393,7 @@ describe("lmstudio-models", () => {
   });
 
   it("reports malformed model list JSON with an owned error", async () => {
-    const fetchMock = vi.fn(async () => ({
-      ok: true,
-      status: 200,
-      json: async () => {
-        throw new SyntaxError("bad json");
-      },
-    }));
+    const fetchMock = vi.fn(async () => malformedJsonResponse());
 
     const result = await fetchLmstudioModels({
       baseUrl: "http://localhost:1234/v1",
@@ -405,11 +406,7 @@ describe("lmstudio-models", () => {
 
   it("reports wrong-shaped model list payloads with owned errors", async () => {
     for (const payload of [[], { models: {} }, { models: [null] }]) {
-      const fetchMock = vi.fn(async () => ({
-        ok: true,
-        status: 200,
-        json: async () => payload,
-      }));
+      const fetchMock = vi.fn(async () => jsonResponse(payload));
 
       const result = await fetchLmstudioModels({
         baseUrl: "http://localhost:1234/v1",
@@ -424,12 +421,9 @@ describe("lmstudio-models", () => {
   it("caps oversized direct fetch timeouts before discovering models", async () => {
     const timeoutController = new AbortController();
     const timeoutSpy = vi.spyOn(AbortSignal, "timeout").mockReturnValue(timeoutController.signal);
-    const fetchMock = vi.fn(async (_url: string | URL, init?: RequestInit) => ({
-      ok: true,
-      status: 200,
-      requestInit: init,
-      json: async () => ({ models: [] }),
-    }));
+    const fetchMock = vi.fn(async (_url: string | URL, _init?: RequestInit) =>
+      jsonResponse({ models: [] }),
+    );
 
     const result = await fetchLmstudioModels({
       baseUrl: "http://localhost:1234/v1",
@@ -521,20 +515,17 @@ describe("lmstudio-models", () => {
     const variantKey = `${canonicalKey}@q4_k_m`;
     const fetchMock = vi.fn(async (url: string | URL) => {
       if (String(url).endsWith("/api/v1/models")) {
-        return {
-          ok: true,
-          json: async () => ({
-            models: [
-              {
-                type: "llm",
-                key: canonicalKey,
-                variants: [variantKey],
-                selected_variant: variantKey,
-                loaded_instances: [],
-              },
-            ],
-          }),
-        };
+        return jsonResponse({
+          models: [
+            {
+              type: "llm",
+              key: canonicalKey,
+              variants: [variantKey],
+              selected_variant: variantKey,
+              loaded_instances: [],
+            },
+          ],
+        });
       }
       if (String(url).endsWith("/api/v1/models/load")) {
         return new Response("load failed", { status: 503 });
@@ -575,20 +566,12 @@ describe("lmstudio-models", () => {
   it("reports malformed model load JSON with an owned error", async () => {
     const fetchMock = vi.fn(async (url: string | URL) => {
       if (String(url).endsWith("/api/v1/models")) {
-        return {
-          ok: true,
-          json: async () => ({
-            models: [{ type: "llm", key: "qwen3-8b-instruct", loaded_instances: [] }],
-          }),
-        };
+        return jsonResponse({
+          models: [{ type: "llm", key: "qwen3-8b-instruct", loaded_instances: [] }],
+        });
       }
       if (String(url).endsWith("/api/v1/models/load")) {
-        return {
-          ok: true,
-          json: async () => {
-            throw new SyntaxError("bad json");
-          },
-        };
+        return malformedJsonResponse();
       }
       throw new Error(`Unexpected fetch URL: ${String(url)}`);
     });
@@ -608,12 +591,9 @@ describe("lmstudio-models", () => {
     const textSpy = vi.spyOn(tracked.response, "text").mockRejectedValue(new Error("unbounded"));
     const fetchMock = vi.fn(async (url: string | URL) => {
       if (String(url).endsWith("/api/v1/models")) {
-        return {
-          ok: true,
-          json: async () => ({
-            models: [{ type: "llm", key: "qwen3-8b-instruct", loaded_instances: [] }],
-          }),
-        };
+        return jsonResponse({
+          models: [{ type: "llm", key: "qwen3-8b-instruct", loaded_instances: [] }],
+        });
       }
       if (String(url).endsWith("/api/v1/models/load")) {
         return tracked.response;

extensions/mattermost/src/mattermost/monitor-gating.test.ts

@@ -59,6 +59,84 @@ describe("mattermost monitor gating", () => {
     });
   });
 
+  it("processes engaged thread follow-ups without a mention", () => {
+    const resolveRequireMention = vi.fn(() => true);
+
+    expect(
+      evaluateMattermostMentionGate({
+        kind: "channel",
+        cfg: {} as never,
+        accountId: "default",
+        channelId: "chan-1",
+        resolveRequireMention,
+        wasMentioned: false,
+        threadAlreadyEngaged: true,
+        isControlCommand: false,
+        commandAuthorized: false,
+        oncharEnabled: false,
+        oncharTriggered: false,
+        canDetectMention: true,
+      }),
+    ).toEqual({
+      shouldRequireMention: true,
+      shouldBypassMention: false,
+      effectiveWasMentioned: true,
+      dropReason: null,
+    });
+  });
+
+  it("engaged threads respond even when onchar is enabled but not triggered", () => {
+    const resolveRequireMention = vi.fn(() => true);
+
+    expect(
+      evaluateMattermostMentionGate({
+        kind: "channel",
+        cfg: {} as never,
+        accountId: "default",
+        channelId: "chan-1",
+        resolveRequireMention,
+        wasMentioned: false,
+        threadAlreadyEngaged: true,
+        isControlCommand: false,
+        commandAuthorized: false,
+        oncharEnabled: true,
+        oncharTriggered: false,
+        canDetectMention: true,
+      }),
+    ).toEqual({
+      shouldRequireMention: true,
+      shouldBypassMention: false,
+      effectiveWasMentioned: true,
+      dropReason: null,
+    });
+  });
+
+  it("drops non-mentioned channel traffic outside an engaged thread", () => {
+    const resolveRequireMention = vi.fn(() => true);
+
+    expect(
+      evaluateMattermostMentionGate({
+        kind: "channel",
+        cfg: {} as never,
+        accountId: "default",
+        channelId: "chan-1",
+        resolveRequireMention,
+        wasMentioned: false,
+        threadAlreadyEngaged: false,
+        isControlCommand: false,
+        commandAuthorized: false,
+        oncharEnabled: false,
+        oncharTriggered: false,
+        canDetectMention: true,
+      }),
+    ).toEqual({
+      shouldRequireMention: true,
+      shouldBypassMention: false,
+      effectiveWasMentioned: false,
+      dropReason: "missing-mention",
+    });
+  });
+
   it("bypasses mention for authorized control commands and allows direct chats", () => {
     const resolveRequireMention = vi.fn(() => true);
 

extensions/mattermost/src/mattermost/monitor-gating.ts

@@ -43,6 +43,9 @@ export type MattermostMentionGateInput = {
   requireMentionOverride?: boolean;
   resolveRequireMention: (params: MattermostRequireMentionResolverInput) => boolean;
   wasMentioned: boolean;
+  // Bot has already replied in this thread; treat follow-ups as addressed so the
+  // user need not re-mention on every turn (parity with Slack thread participation).
+  threadAlreadyEngaged?: boolean;
   isControlCommand: boolean;
   commandAuthorized: boolean;
   oncharEnabled: boolean;
@@ -75,12 +78,16 @@ export function evaluateMattermostMentionGate(
     !params.wasMentioned &&
     params.commandAuthorized;
   const effectiveWasMentioned =
-    params.wasMentioned || shouldBypassMention || params.oncharTriggered;
+    params.wasMentioned ||
+    shouldBypassMention ||
+    params.oncharTriggered ||
+    params.threadAlreadyEngaged === true;
   if (
     params.oncharEnabled &&
     !params.oncharTriggered &&
     !params.wasMentioned &&
-    !params.isControlCommand
+    !params.isControlCommand &&
+    params.threadAlreadyEngaged !== true
   ) {
     return {
       shouldRequireMention,

extensions/mattermost/src/mattermost/monitor.test.ts

@@ -371,6 +371,7 @@ describe("deliverMattermostReplyWithDraftPreview", () => {
   it("suppresses reasoning-prefixed finals before preview finalization", async () => {
     const draftStream = createDraftStreamMock();
     const deliverFinal = vi.fn(async () => {});
+    const recordThreadParticipation = vi.fn();
 
     await deliverMattermostReplyWithDraftPreview({
       payload: { text: "  \n > Reasoning:\n> _hidden_" } as never,
@@ -382,6 +383,7 @@ describe("deliverMattermostReplyWithDraftPreview", () => {
       resolvePreviewFinalText: (text) => text?.trim(),
       previewState: { finalizedViaPreviewPost: false },
       logVerboseMessage: vi.fn(),
+      recordThreadParticipation,
       deliverPayload: deliverFinal,
     });
 
@@ -390,6 +392,36 @@ describe("deliverMattermostReplyWithDraftPreview", () => {
     expect(draftStream.discardPending).not.toHaveBeenCalled();
     expect(draftStream.clear).not.toHaveBeenCalled();
     expect(updateMattermostPostSpy).not.toHaveBeenCalled();
+    // No visible reply was sent, so the thread must not be marked as participated.
+    expect(recordThreadParticipation).not.toHaveBeenCalled();
+  });
+
+  it("records thread participation when a same-thread final finalizes the preview in place", async () => {
+    const draftStream = createDraftStreamMock();
+    const deliverFinal = vi.fn(async () => {});
+    const recordThreadParticipation = vi.fn();
+
+    await deliverMattermostReplyWithDraftPreview({
+      payload: { text: "All good" } as never,
+      info: { kind: "final" },
+      kind: "channel",
+      client: createMattermostClientMock(),
+      draftStream,
+      effectiveReplyToId: "thread-root-1",
+      resolvePreviewFinalText: (text) => text?.trim(),
+      previewState: { finalizedViaPreviewPost: false },
+      logVerboseMessage: vi.fn(),
+      recordThreadParticipation,
+      deliverPayload: deliverFinal,
+    });
+
+    // Default streaming finalizes by editing the preview post, bypassing deliverPayload —
+    // participation must still be recorded (regression: PR #95552 review P1).
+    expect(updateMattermostPostSpy).toHaveBeenCalledWith(expect.anything(), "preview-post-1", {
+      message: "All good",
+    });
+    expect(deliverFinal).not.toHaveBeenCalled();
+    expect(recordThreadParticipation).toHaveBeenCalledTimes(1);
   });
 
   it("deletes the preview after a successful normal final send", async () => {

extensions/mattermost/src/mattermost/monitor.ts

@@ -116,6 +116,10 @@ import {
 import { sendMessageMattermost } from "./send.js";
 import { cleanupSlashCommands } from "./slash-commands.js";
 import { deactivateSlashCommands, getSlashCommandState } from "./slash-state.js";
+import {
+  hasMattermostThreadParticipationWithPersistence,
+  recordMattermostThreadParticipation,
+} from "./thread-participation.js";
 
 export {
   evaluateMattermostMentionGate,
@@ -327,6 +331,10 @@ type MattermostDraftPreviewDeliverParams = {
   previewState: MattermostDraftPreviewState;
   logVerboseMessage: (message: string) => void;
   deliverPayload: (payload: ReplyPayload) => Promise<void>;
+  // Visible same-thread finals can be delivered by editing the draft preview in
+  // place (onPreviewFinalized) without ever calling deliverPayload; this lets the
+  // caller record thread participation on that path too.
+  recordThreadParticipation?: () => void;
 };
 
 export async function deliverMattermostReplyWithDraftPreview(
@@ -374,6 +382,9 @@ export async function deliverMattermostReplyWithDraftPreview(
       },
       onPreviewFinalized: () => {
         params.previewState.finalizedViaPreviewPost = true;
+        // The visible final reply landed by editing the preview post, so the normal
+        // deliverPayload record path is skipped; record participation explicitly here.
+        params.recordThreadParticipation?.();
       },
       buildSupplementalPayload: (payload) =>
         getReplyPayloadTtsSupplement(payload) ? buildTtsSupplementMediaPayload(payload) : undefined,
@@ -1484,6 +1495,16 @@ export async function monitorMattermostProvider(opts: MonitorMattermostOpts = {}
           : { triggered: false, stripped: rawText };
         const oncharTriggered = oncharResult.triggered;
         const canDetectMention = Boolean(botUsername) || mentionRegexes.length > 0;
+        // Threads the bot already replied in auto-engage: follow-ups resume without
+        // a re-mention even under requireMention. Keyed by the thread root id.
+        const threadAlreadyEngaged =
+          kind !== "direct" && effectiveReplyToId
+            ? await hasMattermostThreadParticipationWithPersistence({
+                accountId: account.accountId,
+                channelId,
+                threadRootId: effectiveReplyToId,
+              })
+            : false;
         const mentionDecision = evaluateMattermostMentionGate({
           kind,
           cfg,
@@ -1493,6 +1514,7 @@ export async function monitorMattermostProvider(opts: MonitorMattermostOpts = {}
           requireMentionOverride: account.requireMention,
           resolveRequireMention: core.channel.groups.resolveRequireMention,
           wasMentioned,
+          threadAlreadyEngaged,
           isControlCommand,
           commandAuthorized,
           oncharEnabled,
@@ -1781,6 +1803,18 @@ export async function monitorMattermostProvider(opts: MonitorMattermostOpts = {}
               if (info.kind === "final") {
                 progressDraft.markFinalReplyStarted();
               }
+              // A visible same-thread final arrives either via a normal send or by editing
+              // the draft preview in place; record participation on whichever path fires.
+              const markThreadParticipation = () => {
+                if (kind !== "direct" && effectiveReplyToId) {
+                  recordMattermostThreadParticipation(
+                    account.accountId,
+                    channelId,
+                    effectiveReplyToId,
+                    { agentId: route.agentId },
+                  );
+                }
+              };
               await deliverMattermostReplyWithDraftPreview({
                 payload: payloadEntry,
                 info,
@@ -1791,6 +1825,7 @@ export async function monitorMattermostProvider(opts: MonitorMattermostOpts = {}
                 resolvePreviewFinalText,
                 previewState,
                 logVerboseMessage,
+                recordThreadParticipation: markThreadParticipation,
                 deliverPayload: async (payloadToDeliver) => {
                   const outcome = await deliverMattermostReplyPayload({
                     core,
@@ -1809,6 +1844,11 @@ export async function monitorMattermostProvider(opts: MonitorMattermostOpts = {}
                     sendMessage: sendMessageMattermost,
                     onDmChannelResolution: deliveryBarrier.trackDmChannelResolution,
                   });
+                  // Record only on a visible send so threads we merely observed
+                  // (reasoning-only/empty/suppressed) do not auto-engage later.
+                  if (outcome === "text" || outcome === "media") {
+                    markThreadParticipation();
+                  }
                   const deliveryLog = formatMattermostFinalDeliveryOutcomeLog({
                     outcome,
                     payload: payloadToDeliver,

extensions/mattermost/src/mattermost/thread-participation.test.ts

@@ -0,0 +1,115 @@
+// Mattermost tests cover thread participation cache plugin behavior.
+import type { OpenKeyedStoreOptions } from "openclaw/plugin-sdk/plugin-state-runtime";
+import {
+  createPluginStateKeyedStoreForTests,
+  resetPluginStateStoreForTests,
+} from "openclaw/plugin-sdk/plugin-state-test-runtime";
+import type { PluginRuntime } from "openclaw/plugin-sdk/runtime-store";
+import { afterEach, beforeEach, describe, expect, it } from "vitest";
+import { setMattermostRuntime } from "../runtime.js";
+import {
+  clearMattermostThreadParticipationCache,
+  hasMattermostThreadParticipationWithPersistence,
+  recordMattermostThreadParticipation,
+} from "./thread-participation.js";
+
+// Drain microtasks + the immediate queue so the fire-and-forget persistent write
+// in recordMattermostThreadParticipation has settled before we assert on it.
+const flush = (): Promise<void> =>
+  new Promise((resolve) => {
+    setImmediate(resolve);
+  });
+
+function setRuntime(openKeyedStore: (options: OpenKeyedStoreOptions) => unknown): void {
+  setMattermostRuntime({
+    state: { openKeyedStore },
+    logging: { getChildLogger: () => ({ warn() {} }) },
+  } as unknown as PluginRuntime);
+}
+
+function setPersistentRuntime(): void {
+  setRuntime((options) => createPluginStateKeyedStoreForTests("mattermost", options));
+}
+
+describe("mattermost thread participation", () => {
+  beforeEach(() => {
+    resetPluginStateStoreForTests();
+    clearMattermostThreadParticipationCache();
+    setPersistentRuntime();
+  });
+
+  afterEach(() => {
+    clearMattermostThreadParticipationCache();
+    resetPluginStateStoreForTests();
+  });
+
+  it("remembers a thread the bot replied in", async () => {
+    recordMattermostThreadParticipation("acct", "chan", "root-1");
+    await expect(
+      hasMattermostThreadParticipationWithPersistence({
+        accountId: "acct",
+        channelId: "chan",
+        threadRootId: "root-1",
+      }),
+    ).resolves.toBe(true);
+  });
+
+  it("isolates participation by account, channel, and thread", async () => {
+    recordMattermostThreadParticipation("acct", "chan", "root-1");
+    await flush();
+    for (const probe of [
+      { accountId: "other", channelId: "chan", threadRootId: "root-1" },
+      { accountId: "acct", channelId: "other", threadRootId: "root-1" },
+      { accountId: "acct", channelId: "chan", threadRootId: "root-2" },
+    ]) {
+      await expect(hasMattermostThreadParticipationWithPersistence(probe)).resolves.toBe(false);
+    }
+  });
+
+  it("ignores empty identifiers", async () => {
+    recordMattermostThreadParticipation("", "chan", "root-1");
+    await expect(
+      hasMattermostThreadParticipationWithPersistence({
+        accountId: "",
+        channelId: "chan",
+        threadRootId: "root-1",
+      }),
+    ).resolves.toBe(false);
+  });
+
+  it("recovers participation from the persistent store after the in-memory cache is lost", async () => {
+    recordMattermostThreadParticipation("acct", "chan", "root-1");
+    await flush();
+    // Simulate a restart: in-memory cache cleared, persistent SQLite store intact.
+    clearMattermostThreadParticipationCache();
+    await expect(
+      hasMattermostThreadParticipationWithPersistence({
+        accountId: "acct",
+        channelId: "chan",
+        threadRootId: "root-1",
+      }),
+    ).resolves.toBe(true);
+  });
+
+  it("degrades to in-memory only when the persistent store fails", async () => {
+    setRuntime(() => {
+      throw new Error("sqlite unavailable");
+    });
+    // record + read must not throw; the in-memory cache still answers.
+    recordMattermostThreadParticipation("acct", "chan", "root-1");
+    await expect(
+      hasMattermostThreadParticipationWithPersistence({
+        accountId: "acct",
+        channelId: "chan",
+        threadRootId: "root-1",
+      }),
+    ).resolves.toBe(true);
+    await expect(
+      hasMattermostThreadParticipationWithPersistence({
+        accountId: "acct",
+        channelId: "chan",
+        threadRootId: "missing",
+      }),
+    ).resolves.toBe(false);
+  });
+});

extensions/mattermost/src/mattermost/thread-participation.ts

@@ -0,0 +1,151 @@
+// Mattermost plugin module implements thread participation cache behavior.
+import { resolveGlobalDedupeCache } from "openclaw/plugin-sdk/dedupe-runtime";
+import { getOptionalMattermostRuntime } from "../runtime.js";
+
+/**
+ * In-memory + persisted cache of Mattermost threads the bot has replied in.
+ * Lets the bot auto-respond to thread follow-ups without a re-mention after its
+ * first visible reply. Mirrors the Slack `sent-thread-cache` dual-layer pattern.
+ */
+
+const TTL_MS = 7 * 24 * 60 * 60 * 1000; // 7 days
+const MAX_ENTRIES = 5000;
+const PERSISTENT_MAX_ENTRIES = 1000;
+const PERSISTENT_NAMESPACE = "mattermost.thread-participation";
+
+type MattermostThreadParticipationRecord = {
+  agentId?: string;
+  repliedAt: number;
+};
+
+type MattermostThreadParticipationStore = {
+  register(
+    key: string,
+    value: MattermostThreadParticipationRecord,
+    opts?: { ttlMs?: number },
+  ): Promise<void>;
+  lookup(key: string): Promise<MattermostThreadParticipationRecord | undefined>;
+};
+
+/**
+ * Keep thread participation shared across bundled chunks so thread auto-reply
+ * gating does not diverge between the inbound-gate and reply-dispatch paths.
+ */
+const MATTERMOST_THREAD_PARTICIPATION_KEY = Symbol.for("openclaw.mattermostThreadParticipation");
+const threadParticipation = resolveGlobalDedupeCache(MATTERMOST_THREAD_PARTICIPATION_KEY, {
+  ttlMs: TTL_MS,
+  maxSize: MAX_ENTRIES,
+});
+
+let persistentStore: MattermostThreadParticipationStore | undefined;
+let persistentStoreDisabled = false;
+
+function makeKey(accountId: string, channelId: string, threadRootId: string): string {
+  return `${accountId}:${channelId}:${threadRootId}`;
+}
+
+function reportPersistentThreadParticipationError(error: unknown): void {
+  try {
+    getOptionalMattermostRuntime()
+      ?.logging.getChildLogger({ plugin: "mattermost", feature: "thread-participation-state" })
+      .warn("Mattermost persistent thread participation state failed", { error: String(error) });
+  } catch {
+    // Best effort only: persistent state must never break Mattermost message handling.
+  }
+}
+
+function disablePersistentThreadParticipation(error: unknown): void {
+  persistentStoreDisabled = true;
+  persistentStore = undefined;
+  reportPersistentThreadParticipationError(error);
+}
+
+function getPersistentThreadParticipationStore(): MattermostThreadParticipationStore | undefined {
+  if (persistentStoreDisabled) {
+    return undefined;
+  }
+  if (persistentStore) {
+    return persistentStore;
+  }
+  const runtime = getOptionalMattermostRuntime();
+  if (!runtime) {
+    return undefined;
+  }
+  try {
+    persistentStore = runtime.state.openKeyedStore<MattermostThreadParticipationRecord>({
+      namespace: PERSISTENT_NAMESPACE,
+      maxEntries: PERSISTENT_MAX_ENTRIES,
+      defaultTtlMs: TTL_MS,
+    });
+    return persistentStore;
+  } catch (error) {
+    disablePersistentThreadParticipation(error);
+    return undefined;
+  }
+}
+
+function rememberPersistentThreadParticipation(params: { key: string; agentId?: string }): void {
+  const store = getPersistentThreadParticipationStore();
+  if (!store) {
+    return;
+  }
+  void store
+    .register(params.key, {
+      // Stored for future per-agent thread routing; current reads only need presence.
+      ...(params.agentId ? { agentId: params.agentId } : {}),
+      repliedAt: Date.now(),
+    })
+    .catch(disablePersistentThreadParticipation);
+}
+
+async function lookupPersistentThreadParticipation(key: string): Promise<boolean> {
+  const store = getPersistentThreadParticipationStore();
+  if (!store) {
+    return false;
+  }
+  try {
+    return Boolean(await store.lookup(key));
+  } catch (error) {
+    disablePersistentThreadParticipation(error);
+    return false;
+  }
+}
+
+export function recordMattermostThreadParticipation(
+  accountId: string,
+  channelId: string,
+  threadRootId: string,
+  opts?: { agentId?: string },
+): void {
+  if (!accountId || !channelId || !threadRootId) {
+    return;
+  }
+  const key = makeKey(accountId, channelId, threadRootId);
+  threadParticipation.check(key);
+  rememberPersistentThreadParticipation({ key, agentId: opts?.agentId });
+}
+
+export async function hasMattermostThreadParticipationWithPersistence(params: {
+  accountId: string;
+  channelId: string;
+  threadRootId: string;
+}): Promise<boolean> {
+  if (!params.accountId || !params.channelId || !params.threadRootId) {
+    return false;
+  }
+  const key = makeKey(params.accountId, params.channelId, params.threadRootId);
+  if (threadParticipation.peek(key)) {
+    return true;
+  }
+  const found = await lookupPersistentThreadParticipation(key);
+  if (found) {
+    threadParticipation.check(key);
+  }
+  return found;
+}
+
+export function clearMattermostThreadParticipationCache(): void {
+  threadParticipation.clear();
+  persistentStore = undefined;
+  persistentStoreDisabled = false;
+}

extensions/mattermost/src/runtime.ts

@@ -2,9 +2,12 @@
 import { createPluginRuntimeStore } from "openclaw/plugin-sdk/runtime-store";
 import type { PluginRuntime } from "openclaw/plugin-sdk/runtime-store";
 
-const { setRuntime: setMattermostRuntime, getRuntime: getMattermostRuntime } =
-  createPluginRuntimeStore<PluginRuntime>({
-    pluginId: "mattermost",
-    errorMessage: "Mattermost runtime not initialized",
-  });
-export { getMattermostRuntime, setMattermostRuntime };
+const {
+  setRuntime: setMattermostRuntime,
+  getRuntime: getMattermostRuntime,
+  tryGetRuntime: getOptionalMattermostRuntime,
+} = createPluginRuntimeStore<PluginRuntime>({
+  pluginId: "mattermost",
+  errorMessage: "Mattermost runtime not initialized",
+});
+export { getMattermostRuntime, getOptionalMattermostRuntime, setMattermostRuntime };

extensions/memory-core/src/memory/manager-sync-ops.startup-catchup.test.ts

@@ -3,6 +3,7 @@ import fs from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
 import type { DatabaseSync } from "node:sqlite";
+import { emitSessionTranscriptUpdate } from "openclaw/plugin-sdk/agent-harness-runtime";
 import {
   resolveSessionTranscriptsDirForAgent,
   type OpenClawConfig,
@@ -33,6 +34,25 @@ type SyncParams = {
   progress?: (update: MemorySyncProgressUpdate) => void;
 };
 
+type MemorySessionTranscriptUpdate = {
+  agentId?: string;
+  sessionFile?: string;
+  sessionKey?: string;
+  target?: {
+    agentId: string;
+    sessionId: string;
+    sessionKey: string;
+  };
+};
+
+type MemoryTranscriptUpdateSubscriber = (
+  listener: (update: MemorySessionTranscriptUpdate) => void,
+) => () => void;
+
+const MEMORY_CORE_TRANSCRIPT_UPDATE_SUBSCRIBER_KEY = Symbol.for(
+  "openclaw.memoryCore.sessionTranscriptUpdateSubscriber",
+);
+
 type SourceStateRow = { path: string; hash: string; mtime: number; size: number };
 
 class SessionStartupCatchupHarness extends MemoryManagerSyncOps {
@@ -111,10 +131,27 @@ class SessionStartupCatchupHarness extends MemoryManagerSyncOps {
     return Array.from(this.sessionsDirtyFiles);
   }
 
+  getPendingSessionTargets(): MemorySyncParams["sessions"] {
+    return Array.from(this.sessionPendingTargets.values());
+  }
+
+  getPendingSessionFiles(): string[] {
+    return Array.from(this.sessionPendingFiles);
+  }
+
   isSessionsDirty(): boolean {
     return this.sessionsDirty;
   }
 
+  startTranscriptListener(): void {
+    this.ensureSessionListener();
+  }
+
+  stopTranscriptListener(): void {
+    this.sessionUnsubscribe?.();
+    this.sessionUnsubscribe = null;
+  }
+
   protected computeProviderKey(): string {
     return "test";
   }
@@ -162,6 +199,8 @@ describe("session startup catch-up", () => {
   });
 
   afterEach(async () => {
+    vi.clearAllTimers();
+    vi.useRealTimers();
     vi.unstubAllEnvs();
     await fs.rm(stateDir, { recursive: true, force: true });
   });
@@ -356,4 +395,84 @@ describe("session startup catch-up", () => {
 
     expect(harness.indexedPaths).toEqual([]);
   });
+
+  it("queues transcript update identity without requiring a session file", async () => {
+    vi.useFakeTimers();
+    const harness = new SessionStartupCatchupHarness([]);
+    const originalSubscriber = (globalThis as Record<symbol, unknown>)[
+      MEMORY_CORE_TRANSCRIPT_UPDATE_SUBSCRIBER_KEY
+    ];
+    let transcriptListener: ((update: MemorySessionTranscriptUpdate) => void) | undefined;
+    (globalThis as Record<symbol, unknown>)[MEMORY_CORE_TRANSCRIPT_UPDATE_SUBSCRIBER_KEY] = ((
+      listener,
+    ) => {
+      transcriptListener = listener;
+      return () => {
+        if (transcriptListener === listener) {
+          transcriptListener = undefined;
+        }
+      };
+    }) satisfies MemoryTranscriptUpdateSubscriber;
+    harness.startTranscriptListener();
+
+    try {
+      transcriptListener?.({
+        target: {
+          agentId: "main",
+          sessionId: "thread",
+          sessionKey: "agent:main:thread",
+        },
+      });
+
+      expect(harness.getPendingSessionTargets()).toEqual([
+        { agentId: "main", sessionId: "thread", sessionKey: "agent:main:thread" },
+      ]);
+    } finally {
+      harness.stopTranscriptListener();
+      if (originalSubscriber === undefined) {
+        delete (globalThis as Record<symbol, unknown>)[
+          MEMORY_CORE_TRANSCRIPT_UPDATE_SUBSCRIBER_KEY
+        ];
+      } else {
+        (globalThis as Record<symbol, unknown>)[MEMORY_CORE_TRANSCRIPT_UPDATE_SUBSCRIBER_KEY] =
+          originalSubscriber;
+      }
+    }
+  });
+
+  it("keeps canonical path transcript update compatibility", async () => {
+    vi.useFakeTimers();
+    const session = await writeSessionFile("thread.jsonl");
+    const harness = new SessionStartupCatchupHarness([]);
+    harness.startTranscriptListener();
+
+    emitSessionTranscriptUpdate({
+      sessionFile: session.filePath,
+      sessionKey: "agent:main:thread",
+    });
+
+    expect(harness.getPendingSessionFiles()).toEqual([session.filePath]);
+    expect(harness.getPendingSessionTargets()).toEqual([]);
+    harness.stopTranscriptListener();
+  });
+
+  it("prefers transcript update path compatibility before identity", async () => {
+    vi.useFakeTimers();
+    const session = await writeSessionFile("thread.jsonl");
+    const harness = new SessionStartupCatchupHarness([]);
+    harness.startTranscriptListener();
+
+    emitSessionTranscriptUpdate({
+      sessionFile: session.filePath,
+      target: {
+        agentId: "main",
+        sessionId: "identity-target",
+        sessionKey: "agent:main:identity-target",
+      },
+    });
+
+    expect(harness.getPendingSessionFiles()).toEqual([session.filePath]);
+    expect(harness.getPendingSessionTargets()).toEqual([]);
+    harness.stopTranscriptListener();
+  });
 });

extensions/memory-core/src/memory/manager-sync-ops.ts

@@ -170,9 +170,27 @@ const IGNORED_MEMORY_WATCH_DIR_NAMES = new Set([
 ]);
 
 const log = createSubsystemLogger("memory");
+const MEMORY_CORE_TRANSCRIPT_UPDATE_SUBSCRIBER_KEY = Symbol.for(
+  "openclaw.memoryCore.sessionTranscriptUpdateSubscriber",
+);
 const TEST_MEMORY_WATCH_FACTORY_KEY = Symbol.for("openclaw.test.memoryWatchFactory");
 const TEST_MEMORY_NATIVE_WATCH_FACTORY_KEY = Symbol.for("openclaw.test.memoryNativeWatchFactory");
 
+type MemorySessionTranscriptUpdate = {
+  agentId?: string;
+  sessionFile?: string;
+  sessionKey?: string;
+  target?: {
+    agentId: string;
+    sessionId: string;
+    sessionKey: string;
+  };
+};
+
+type MemoryTranscriptUpdateSubscriber = (
+  listener: (update: MemorySessionTranscriptUpdate) => void,
+) => () => void;
+
 function memoryTableExists(db: DatabaseSync, tableName: string): boolean {
   return Boolean(
     db.prepare("SELECT 1 FROM sqlite_master WHERE type = 'table' AND name = ?").get(tableName),
@@ -191,6 +209,18 @@ type LinuxMemoryDirectoryWatcher = {
   ino: number;
 };
 
+function subscribeMemorySessionTranscriptUpdates(
+  listener: (update: MemorySessionTranscriptUpdate) => void,
+): () => void {
+  const injected = (globalThis as Record<symbol, unknown>)[
+    MEMORY_CORE_TRANSCRIPT_UPDATE_SUBSCRIBER_KEY
+  ];
+  if (typeof injected === "function") {
+    return (injected as MemoryTranscriptUpdateSubscriber)(listener);
+  }
+  return onSessionTranscriptUpdate(listener);
+}
+
 function resolveMemoryWatchFactory(): typeof chokidar.watch {
   if (process.env.VITEST === "true" || process.env.NODE_ENV === "test") {
     const override = (globalThis as Record<PropertyKey, unknown>)[TEST_MEMORY_WATCH_FACTORY_KEY];
@@ -1422,20 +1452,22 @@ export abstract class MemoryManagerSyncOps {
     if (!this.sources.has("sessions") || this.sessionUnsubscribe) {
       return;
     }
-    this.sessionUnsubscribe = onSessionTranscriptUpdate((update) => {
+    this.sessionUnsubscribe = subscribeMemorySessionTranscriptUpdates((update) => {
       if (this.closed) {
         return;
       }
       const sessionFile = update.sessionFile;
-      if (!this.isSessionFileForAgent(sessionFile)) {
+      if (sessionFile && isSessionArchiveArtifactName(path.basename(sessionFile))) {
+        return;
+      }
+      if (sessionFile && this.isSessionFileForAgent(sessionFile)) {
+        this.scheduleSessionDirty(sessionFile);
         return;
       }
       const target = this.resolveSessionTranscriptUpdateSyncTarget(update);
       if (target) {
         this.scheduleSessionDirty(target);
-        return;
       }
-      this.scheduleSessionDirty(sessionFile);
     });
   }
 
@@ -1703,13 +1735,30 @@ export abstract class MemoryManagerSyncOps {
     return resolvedFile.startsWith(`${resolvedDir}${path.sep}`);
   }
 
-  private resolveSessionTranscriptUpdateSyncTarget(update: {
-    agentId?: string;
-    sessionFile: string;
-    sessionKey?: string;
-  }): MemorySessionSyncTarget | null {
+  private resolveSessionTranscriptUpdateSyncTarget(
+    update: MemorySessionTranscriptUpdate,
+  ): MemorySessionSyncTarget | null {
+    if (update.sessionFile && isSessionArchiveArtifactName(path.basename(update.sessionFile))) {
+      return null;
+    }
+    if (update.target) {
+      const agentId = update.target.agentId.trim();
+      const sessionId = update.target.sessionId.trim();
+      const sessionKey = update.target.sessionKey.trim();
+      if (!agentId || !sessionId || normalizeAgentId(agentId) !== normalizeAgentId(this.agentId)) {
+        return null;
+      }
+      return {
+        agentId,
+        sessionId,
+        ...(sessionKey ? { sessionKey } : {}),
+      };
+    }
+    if (!update.sessionFile) {
+      return null;
+    }
     const parsed = parseCanonicalSessionSyncTargetFromPath(update.sessionFile);
-    if (!parsed || isSessionArchiveArtifactName(path.basename(update.sessionFile))) {
+    if (!parsed) {
       return null;
     }
     const agentId = update.agentId?.trim() || parsed.agentId;

extensions/memory-wiki/src/compile.test.ts

@@ -559,6 +559,85 @@ describe("compileMemoryWikiVault", () => {
     );
   });
 
+  it("excludes concept and synthesis pages from stale-pages report", async () => {
+    const { rootDir, config } = await createVault({
+      rootDir: nextCaseRoot(),
+      initialize: true,
+    });
+
+    await fs.writeFile(
+      path.join(rootDir, "entities", "entity-alpha.md"),
+      renderWikiMarkdown({
+        frontmatter: {
+          pageType: "entity",
+          id: "entity.alpha",
+          title: "Alpha Entity",
+          sourceIds: ["source.alpha"],
+          updatedAt: "2025-06-01T00:00:00.000Z",
+        },
+        body: "# Alpha Entity\n",
+      }),
+      "utf8",
+    );
+
+    await fs.writeFile(
+      path.join(rootDir, "sources", "source-alpha.md"),
+      renderWikiMarkdown({
+        frontmatter: {
+          pageType: "source",
+          id: "source.alpha",
+          title: "Alpha Source",
+          updatedAt: "2025-06-01T00:00:00.000Z",
+        },
+        body: "# Alpha Source\n",
+      }),
+      "utf8",
+    );
+
+    // Concept page with old updatedAt — should be excluded from stale-pages
+    await fs.writeFile(
+      path.join(rootDir, "concepts", "concept-beta.md"),
+      renderWikiMarkdown({
+        frontmatter: {
+          pageType: "concept",
+          id: "concept.beta",
+          title: "Beta Concept",
+          sourceIds: ["source.alpha"],
+          updatedAt: "2025-06-01T00:00:00.000Z",
+        },
+        body: "# Beta Concept\n",
+      }),
+      "utf8",
+    );
+
+    // Synthesis page with old updatedAt — should be excluded from stale-pages
+    await fs.writeFile(
+      path.join(rootDir, "syntheses", "synthesis-gamma.md"),
+      renderWikiMarkdown({
+        frontmatter: {
+          pageType: "synthesis",
+          id: "synthesis.gamma",
+          title: "Gamma Synthesis",
+          sourceIds: ["source.alpha"],
+          updatedAt: "2025-06-01T00:00:00.000Z",
+        },
+        body: "# Gamma Synthesis\n",
+      }),
+      "utf8",
+    );
+
+    await compileMemoryWikiVault(config);
+
+    const stalePages = await fs.readFile(path.join(rootDir, "reports", "stale-pages.md"), "utf8");
+
+    // Entity and source pages still appear in stale-pages
+    expect(stalePages).toContain("[Alpha Entity](../entities/entity-alpha.md)");
+    expect(stalePages).toContain("[Alpha Source](../sources/source-alpha.md)");
+    // Concept and synthesis pages are excluded
+    expect(stalePages).not.toContain("[Beta Concept](../concepts/concept-beta.md)");
+    expect(stalePages).not.toContain("[Gamma Synthesis](../syntheses/synthesis-gamma.md)");
+  });
+
   it("skips dashboard report pages when createDashboards is disabled", async () => {
     const { rootDir, config } = await createVault({
       rootDir: nextCaseRoot(),

extensions/memory-wiki/src/compile.ts

@@ -214,6 +214,9 @@ const DASHBOARD_PAGES: DashboardPageDefinition[] = [
         .filter(
           (page) =>
             page.kind !== "report" &&
+            // concept/synthesis are intentionally durable references
+            page.kind !== "concept" &&
+            page.kind !== "synthesis" &&
             !(
               isUnmanagedRawSourceSummary(page) &&
               !managedImportedSourcePagePaths.has(page.relativePath)

extensions/nextcloud-talk/src/room-info.test.ts

@@ -36,20 +36,25 @@ function requireFirstFetchParams(): {
   return fetchParams as { auditContext?: string; url?: string };
 }
 
+function jsonResponse(payload: unknown, init?: ResponseInit): Response {
+  return new Response(JSON.stringify(payload), {
+    status: 200,
+    headers: { "content-type": "application/json" },
+    ...init,
+  });
+}
+
 describe("nextcloud talk room info", () => {
   it("resolves direct rooms from the room info endpoint", async () => {
     const release = vi.fn(async () => {});
     fetchWithSsrFGuard.mockResolvedValue({
-      response: {
-        ok: true,
-        json: async () => ({
-          ocs: {
-            data: {
-              type: 1,
-            },
+      response: jsonResponse({
+        ocs: {
+          data: {
+            type: 1,
           },
-        }),
-      },
+        },
+      }),
       release,
     });
 
@@ -76,16 +81,13 @@ describe("nextcloud talk room info", () => {
 
   it("normalizes signed decimal room type strings through the shared parser", async () => {
     fetchWithSsrFGuard.mockResolvedValue({
-      response: {
-        ok: true,
-        json: async () => ({
-          ocs: {
-            data: {
-              type: "+01",
-            },
+      response: jsonResponse({
+        ocs: {
+          data: {
+            type: "+01",
           },
-        }),
-      },
+        },
+      }),
       release: vi.fn(async () => {}),
     });
 
@@ -106,16 +108,13 @@ describe("nextcloud talk room info", () => {
 
   it("does not coerce partial room type strings", async () => {
     fetchWithSsrFGuard.mockResolvedValue({
-      response: {
-        ok: true,
-        json: async () => ({
-          ocs: {
-            data: {
-              type: "1direct",
-            },
+      response: jsonResponse({
+        ocs: {
+          data: {
+            type: "1direct",
           },
-        }),
-      },
+        },
+      }),
       release: vi.fn(async () => {}),
     });
 
@@ -136,16 +135,13 @@ describe("nextcloud talk room info", () => {
 
   it("does not classify negative room types as group rooms", async () => {
     fetchWithSsrFGuard.mockResolvedValue({
-      response: {
-        ok: true,
-        json: async () => ({
-          ocs: {
-            data: {
-              type: -1,
-            },
+      response: jsonResponse({
+        ocs: {
+          data: {
+            type: -1,
           },
-        }),
-      },
+        },
+      }),
       release: vi.fn(async () => {}),
     });
 

extensions/qa-lab/package.json

@@ -16,7 +16,7 @@
     "@openclaw/plugin-sdk": "workspace:*",
     "@openclaw/slack": "workspace:*",
     "@openclaw/whatsapp": "workspace:*",
-    "crabline": "github:openclaw/crabline#b3513f66053788c6a7bd2bc76fbfc7201f647d29",
+    "@openclaw/crabline": "0.1.0",
     "openclaw": "2026.5.28"
   },
   "peerDependencies": {

extensions/qa-lab/src/artifact-run-id.ts

@@ -0,0 +1,6 @@
+// Qa Lab plugin helper creates collision-resistant artifact run identifiers.
+import { randomUUID } from "node:crypto";
+
+export function createQaArtifactRunId(): string {
+  return `${Date.now().toString(36)}-${randomUUID().slice(0, 8)}`;
+}

extensions/qa-lab/src/character-eval.test.ts

@@ -255,6 +255,39 @@ describe("runQaCharacterEval", () => {
     expect(report).not.toContain("Judge Raw Reply");
   });
 
+  it("creates a unique default output directory under repo artifacts", async () => {
+    const runSuite = vi.fn(async (params: CharacterRunSuiteParams) =>
+      makeSuiteResult({
+        outputDir: params.outputDir,
+        model: params.primaryModel,
+        transcript: "USER Alice: hi\n\nASSISTANT openclaw: default dir reply",
+      }),
+    );
+    const runJudge = makeRunJudge([
+      {
+        model: "openai/gpt-5.5",
+        rank: 1,
+        score: 8,
+        summary: "solid",
+        strengths: ["clear"],
+        weaknesses: [],
+      },
+    ]);
+
+    const result = await runQaCharacterEval({
+      repoRoot: tempRoot,
+      models: ["openai/gpt-5.5"],
+      runSuite,
+      runJudge,
+    });
+
+    expect(path.dirname(result.outputDir)).toBe(path.join(tempRoot, ".artifacts", "qa-e2e"));
+    expect(path.basename(result.outputDir)).toMatch(
+      /^character-eval-[a-z0-9]+-[a-f0-9]{8}$/u,
+    );
+    await expect(fs.stat(result.reportPath).then((stats) => stats.isFile())).resolves.toBe(true);
+  });
+
   it("can hide candidate model refs from judge prompts and map rankings back", async () => {
     const runSuite = vi.fn(async (params: CharacterRunSuiteParams) =>
       makeSuiteResult({

extensions/qa-lab/src/character-eval.ts

@@ -3,6 +3,7 @@ import fs from "node:fs/promises";
 import path from "node:path";
 import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { normalizeStringEntries, uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";
+import { createQaArtifactRunId } from "./artifact-run-id.js";
 import { isQaFastModeModelRef, type QaProviderMode } from "./model-selection.js";
 import {
   QA_FRONTIER_CHARACTER_EVAL_MODELS,
@@ -520,7 +521,7 @@ export async function runQaCharacterEval(params: QaCharacterEvalParams) {
 
   const outputDir =
     params.outputDir ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `character-eval-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `character-eval-${createQaArtifactRunId()}`);
   const runsDir = path.join(outputDir, "runs");
   await fs.mkdir(runsDir, { recursive: true });
 

extensions/qa-lab/src/cli.runtime.ts

@@ -4,7 +4,7 @@ import path from "node:path";
 import {
   OPENCLAW_CRABLINE_DEFAULT_CHANNEL,
   resolveOpenClawCrablineChannelDriverSelection,
-} from "crabline";
+} from "@openclaw/crabline";
 import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { parseStrictPositiveInteger } from "openclaw/plugin-sdk/number-runtime";
 import { uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";
@@ -17,6 +17,7 @@ import {
   type QaRuntimeParitySuiteSummary,
 } from "./agentic-parity-report.js";
 import { resolveQaParityPackScenarioIds } from "./agentic-parity.js";
+import { createQaArtifactRunId } from "./artifact-run-id.js";
 import { runQaCharacterEval, type QaCharacterModelOptions } from "./character-eval.js";
 import { resolveRepoRelativeOutputDir } from "./cli-paths.js";
 import {
@@ -407,7 +408,7 @@ async function runQaParityPreflight(params: {
     ".artifacts",
     "qa-e2e",
     "preflight",
-    `suite-${Date.now().toString(36)}`,
+    `suite-${createQaArtifactRunId()}`,
   );
   const result = await runQaSuiteWithInfraRetry(() =>
     runQaFlowSuiteFromRuntime({
@@ -1056,7 +1057,7 @@ export async function runQaParityReportCommand(opts: {
   }
   const outputDir =
     resolveRepoRelativeOutputDir(repoRoot, opts.outputDir) ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `parity-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `parity-${createQaArtifactRunId()}`);
   await fs.mkdir(outputDir, { recursive: true });
 
   if (opts.runtimeAxis === true) {
@@ -1149,7 +1150,7 @@ export async function runQaConfidenceReportCommand(opts: {
   const artifactRoot = path.resolve(repoRoot, opts.artifactRoot ?? ".");
   const outputDir =
     resolveRepoRelativeOutputDir(repoRoot, opts.outputDir) ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `confidence-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `confidence-${createQaArtifactRunId()}`);
   await fs.mkdir(outputDir, { recursive: true });
   const manifest = await readQaConfidenceManifestFile(manifestPath);
   const reportPayload = await buildQaConfidenceReport({
@@ -1178,7 +1179,7 @@ export async function runQaConfidenceSelfTestCommand(opts: {
   const repoRoot = path.resolve(opts.repoRoot ?? process.cwd());
   const outputDir =
     resolveRepoRelativeOutputDir(repoRoot, opts.outputDir) ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `confidence-self-test-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `confidence-self-test-${createQaArtifactRunId()}`);
   const result = await writeQaConfidenceSelfTestArtifacts({ outputDir });
   process.stdout.write(`QA confidence self-test report: ${result.reportPath}\n`);
   process.stdout.write(`QA confidence self-test summary: ${result.summaryPath}\n`);
@@ -1268,7 +1269,7 @@ export async function runQaJsonlReplayCommand(opts: {
   const transcriptDir = path.resolve(repoRoot, opts.transcripts ?? "qa/scenarios/jsonl-replay");
   const outputDir =
     resolveRepoRelativeOutputDir(repoRoot, opts.outputDir) ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `jsonl-replay-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `jsonl-replay-${createQaArtifactRunId()}`);
   await fs.mkdir(outputDir, { recursive: true });
   const result = await runJsonlReplay(
     {

extensions/qa-lab/src/cli.test.ts

@@ -800,6 +800,33 @@ describe("qa cli registration", () => {
     await expect(invalidProgram.parseAsync(["node", "openclaw", ...args])).rejects.toThrow(message);
   });
 
+  it.each([
+    [["qa", "ui", "--port", "65536"], "--port must be a TCP port between 1 and 65535."],
+    [
+      ["qa", "ui", "--advertise-port", "999999"],
+      "--advertise-port must be a TCP port between 1 and 65535.",
+    ],
+    [
+      ["qa", "docker-scaffold", "--output-dir", "/tmp/qa", "--gateway-port", "65536"],
+      "--gateway-port must be a TCP port between 1 and 65535.",
+    ],
+    [
+      ["qa", "up", "--qa-lab-port", "65536"],
+      "--qa-lab-port must be a TCP port between 1 and 65535.",
+    ],
+    [["qa", "aimock", "--port", "65536"], "--port must be a TCP port between 1 and 65535."],
+  ])("rejects out-of-range QA port option %j", async (args, message) => {
+    const invalidProgram = new Command();
+    invalidProgram.exitOverride();
+    invalidProgram.configureOutput({
+      writeErr: () => {},
+      writeOut: () => {},
+    });
+    registerQaLabCli(invalidProgram);
+
+    await expect(invalidProgram.parseAsync(["node", "openclaw", ...args])).rejects.toThrow(message);
+  });
+
   it("shows an enable hint when a discovered runner plugin is installed but blocked", async () => {
     listQaRunnerCliContributions.mockReset().mockReturnValue([createBlockedQaRunnerContribution()]);
     const blockedProgram = new Command();

extensions/qa-lab/src/cli.ts

@@ -62,6 +62,7 @@ const QA_RUN_PROFILE_ONLY_OPTIONS = [
 ] as const;
 
 const QA_RUN_SELF_CHECK_ONLY_OPTIONS = [{ optionName: "output", flag: "--output" }] as const;
+const MAX_QA_CLI_TCP_PORT = 65_535;
 
 type QaSuiteCliOptions = QaScenarioRunCliOptions & {
   channelDriver?: QaSuiteCommandOptions["channelDriver"];
@@ -105,6 +106,14 @@ function parseQaCliPositiveIntegerOption(value: string, flag: string): number {
   return parsed;
 }
 
+function parseQaCliTcpPortOption(value: string, flag: string): number {
+  const parsed = parseQaCliPositiveIntegerOption(value, flag);
+  if (parsed > MAX_QA_CLI_TCP_PORT) {
+    throw invalidQaCliArgument(`${flag} must be a TCP port between 1 and 65535.`);
+  }
+  return parsed;
+}
+
 function parseQaEvidenceModeOption(value: string): QaProfileCommandOptions["evidenceMode"] {
   const evidenceMode = value.trim();
   if (evidenceMode === "full" || evidenceMode === "slim") {
@@ -867,11 +876,11 @@ export function registerQaLabCli(program: Command) {
     .option("--repo-root <path>", "Repository root to target when running from a neutral cwd")
     .option("--host <host>", "Bind host", "127.0.0.1")
     .option("--port <port>", "Bind port", (value: string) =>
-      parseQaCliPositiveIntegerOption(value, "--port"),
+      parseQaCliTcpPortOption(value, "--port"),
     )
     .option("--advertise-host <host>", "Optional public host to advertise in bootstrap payloads")
     .option("--advertise-port <port>", "Optional public port to advertise", (value: string) =>
-      parseQaCliPositiveIntegerOption(value, "--advertise-port"),
+      parseQaCliTcpPortOption(value, "--advertise-port"),
     )
     .option("--control-ui-url <url>", "Optional Control UI URL to embed beside the QA panel")
     .option(
@@ -909,10 +918,10 @@ export function registerQaLabCli(program: Command) {
     .option("--repo-root <path>", "Repository root to target when running from a neutral cwd")
     .requiredOption("--output-dir <path>", "Output directory for docker-compose + state files")
     .option("--gateway-port <port>", "Gateway host port", (value: string) =>
-      parseQaCliPositiveIntegerOption(value, "--gateway-port"),
+      parseQaCliTcpPortOption(value, "--gateway-port"),
     )
     .option("--qa-lab-port <port>", "QA lab host port", (value: string) =>
-      parseQaCliPositiveIntegerOption(value, "--qa-lab-port"),
+      parseQaCliTcpPortOption(value, "--qa-lab-port"),
     )
     .option("--provider-base-url <url>", "Provider base URL for the QA gateway")
     .option("--image <name>", "Prebaked image name", "openclaw:qa-local-prebaked")
@@ -950,10 +959,10 @@ export function registerQaLabCli(program: Command) {
     .option("--repo-root <path>", "Repository root to target when running from a neutral cwd")
     .option("--output-dir <path>", "Output directory for docker-compose + state files")
     .option("--gateway-port <port>", "Gateway host port", (value: string) =>
-      parseQaCliPositiveIntegerOption(value, "--gateway-port"),
+      parseQaCliTcpPortOption(value, "--gateway-port"),
     )
     .option("--qa-lab-port <port>", "QA lab host port", (value: string) =>
-      parseQaCliPositiveIntegerOption(value, "--qa-lab-port"),
+      parseQaCliTcpPortOption(value, "--qa-lab-port"),
     )
     .option("--provider-base-url <url>", "Provider base URL for the QA gateway")
     .option("--image <name>", "Image tag", "openclaw:qa-local-prebaked")
@@ -985,7 +994,7 @@ export function registerQaLabCli(program: Command) {
       .description(providerCommand.description)
       .option("--host <host>", "Bind host", "127.0.0.1")
       .option("--port <port>", "Bind port", (value: string) =>
-        parseQaCliPositiveIntegerOption(value, "--port"),
+        parseQaCliTcpPortOption(value, "--port"),
       )
       .action(async (opts: { host?: string; port?: number }) => {
         await runQaProviderServer(providerCommand.providerMode, opts);

extensions/qa-lab/src/crabline-transport.test.ts

@@ -1,7 +1,7 @@
 // Qa Lab tests cover Crabline fake-provider transport integration behavior.
 import fs from "node:fs/promises";
 import path from "node:path";
-import { OPENCLAW_CRABLINE_MANIFEST_PATH } from "crabline";
+import { OPENCLAW_CRABLINE_MANIFEST_PATH } from "@openclaw/crabline";
 import { fetchWithSsrFGuard } from "openclaw/plugin-sdk/ssrf-runtime";
 import { withTempDir } from "openclaw/plugin-sdk/test-env";
 import { describe, expect, it } from "vitest";

extensions/qa-lab/src/crabline-transport.ts

@@ -7,7 +7,7 @@ import {
   startOpenClawCrablineAdapter,
   type OpenClawCrablineChannelDriverSelection,
   type StartedOpenClawCrablineAdapter,
-} from "crabline";
+} from "@openclaw/crabline";
 import type { OpenClawConfig } from "openclaw/plugin-sdk/config-contracts";
 import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { fetchWithSsrFGuard } from "openclaw/plugin-sdk/ssrf-runtime";
@@ -101,7 +101,10 @@ async function postCrablineInbound(params: {
     url: params.adapter.manifest.endpoints.adminInboundUrl,
     init: {
       body: JSON.stringify(params.providerBody),
-      headers: { "content-type": "application/json" },
+      headers: {
+        authorization: `Bearer ${params.adapter.manifest.adminToken}`,
+        "content-type": "application/json",
+      },
       method: "POST",
     },
     policy: { allowPrivateNetwork: true },

extensions/qa-lab/src/gateway-child.test.ts

@@ -1231,6 +1231,9 @@ describe("buildQaRuntimeEnv", () => {
     await expect(readFile(path.join(artifactDir, "README.txt"), "utf8")).resolves.toContain(
       "was not copied because it may contain credentials or auth tokens",
     );
+    await expect(readFile(path.join(artifactDir, "README.txt"), "utf8")).resolves.not.toContain(
+      tempRoot,
+    );
   });
 
   it("rejects preserved gateway artifacts outside the repo root", async () => {

extensions/qa-lab/src/gateway-child.ts

@@ -162,7 +162,7 @@ async function preserveQaGatewayDebugArtifacts(params: {
     [
       "Only sanitized gateway debug artifacts are preserved here.",
       "The full QA gateway runtime was not copied because it may contain credentials or auth tokens.",
-      `Original runtime temp root: ${params.tempRoot}`,
+      "Original runtime temp root omitted because local temp paths can identify the runner.",
       "",
     ].join("\n"),
     "utf8",

extensions/qa-lab/src/lab-server.test.ts

@@ -657,7 +657,8 @@ describe("qa-lab server", () => {
     });
 
     const result = await lab.runSelfCheck();
-    expect(result.outputPath).toBe(path.join(repoRoot, ".artifacts", "qa-e2e", "self-check.md"));
+    expect(path.dirname(result.outputPath)).toBe(path.join(repoRoot, ".artifacts", "qa-e2e"));
+    expect(path.basename(result.outputPath)).toMatch(/^self-check-[a-z0-9]+-[a-f0-9]{8}\.md$/u);
     expect(await readFile(result.outputPath, "utf8")).toContain("Synthetic Slack-class roundtrip");
   });
 

extensions/qa-lab/src/live-transports/discord/discord-live.runtime.ts

@@ -15,6 +15,7 @@ import { writeExternalFileWithinRoot } from "openclaw/plugin-sdk/security-runtim
 import { uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { chromium } from "playwright-core";
 import { z } from "zod";
+import { createQaArtifactRunId } from "../../artifact-run-id.js";
 import { QA_EVIDENCE_FILENAME, buildLiveTransportEvidenceSummary } from "../../evidence-summary.js";
 import { startQaGatewayChild } from "../../gateway-child.js";
 import { isTruthyOptIn } from "../../mantis-options.runtime.js";
@@ -1527,7 +1528,7 @@ export async function runDiscordQaLive(params: {
   const repoRoot = path.resolve(params.repoRoot ?? process.cwd());
   const outputDir =
     params.outputDir ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `discord-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `discord-${createQaArtifactRunId()}`);
   await fs.mkdir(outputDir, { recursive: true });
 
   const providerMode = normalizeQaProviderMode(

extensions/qa-lab/src/live-transports/slack/slack-live.runtime.ts

@@ -9,6 +9,7 @@ import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { parseStrictPositiveInteger } from "openclaw/plugin-sdk/number-runtime";
 import { uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { z } from "zod";
+import { createQaArtifactRunId } from "../../artifact-run-id.js";
 import { QA_EVIDENCE_FILENAME, buildLiveTransportEvidenceSummary } from "../../evidence-summary.js";
 import { startQaGatewayChild } from "../../gateway-child.js";
 import { isTruthyOptIn } from "../../mantis-options.runtime.js";
@@ -1711,7 +1712,7 @@ export async function runSlackQaLive(params: {
   const repoRoot = path.resolve(params.repoRoot ?? process.cwd());
   const outputDir =
     params.outputDir ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `slack-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `slack-${createQaArtifactRunId()}`);
   await fs.mkdir(outputDir, { recursive: true });
 
   const providerMode = normalizeQaProviderMode(

extensions/qa-lab/src/live-transports/telegram/telegram-live.runtime.ts

@@ -11,6 +11,7 @@ import {
 import { fetchWithSsrFGuard } from "openclaw/plugin-sdk/ssrf-runtime";
 import { isRecord, uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { z } from "zod";
+import { createQaArtifactRunId } from "../../artifact-run-id.js";
 import {
   QA_EVIDENCE_FILENAME,
   buildLiveTransportEvidenceSummary,
@@ -1805,7 +1806,7 @@ export async function runTelegramQaLive(params: {
   const repoRoot = path.resolve(params.repoRoot ?? process.cwd());
   const outputDir =
     params.outputDir ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `telegram-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `telegram-${createQaArtifactRunId()}`);
   await fs.mkdir(outputDir, { recursive: true });
 
   const providerMode = normalizeQaProviderMode(

extensions/qa-lab/src/live-transports/whatsapp/whatsapp-live.runtime.ts

@@ -15,6 +15,7 @@ import { formatErrorMessage } from "openclaw/plugin-sdk/error-runtime";
 import { normalizeStringEntries, uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { resolvePreferredOpenClawTmpDir } from "openclaw/plugin-sdk/temp-path";
 import { z } from "zod";
+import { createQaArtifactRunId } from "../../artifact-run-id.js";
 import { QA_EVIDENCE_FILENAME, buildLiveTransportEvidenceSummary } from "../../evidence-summary.js";
 import { startQaGatewayChild } from "../../gateway-child.js";
 import { isTruthyOptIn } from "../../mantis-options.runtime.js";
@@ -3033,7 +3034,7 @@ export async function runWhatsAppQaLive(params: {
   const repoRoot = path.resolve(params.repoRoot ?? process.cwd());
   const outputDir =
     params.outputDir ??
-    path.join(repoRoot, ".artifacts", "qa-e2e", `whatsapp-${Date.now().toString(36)}`);
+    path.join(repoRoot, ".artifacts", "qa-e2e", `whatsapp-${createQaArtifactRunId()}`);
   await fs.mkdir(outputDir, { recursive: true });
 
   const providerMode = normalizeQaProviderMode(

extensions/qa-lab/src/multipass.runtime.ts

@@ -4,7 +4,7 @@ import { randomUUID } from "node:crypto";
 import fs from "node:fs";
 import { access, mkdir, writeFile } from "node:fs/promises";
 import path from "node:path";
-import type { OpenClawCrablineChannelDriverSelection } from "crabline";
+import type { OpenClawCrablineChannelDriverSelection } from "@openclaw/crabline";
 import { sleep } from "openclaw/plugin-sdk/runtime-env";
 import { appendRegularFile } from "openclaw/plugin-sdk/security-runtime";
 import { uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";

extensions/qa-lab/src/providers/mock-openai/server.test.ts

@@ -3314,7 +3314,7 @@ describe("qa mock openai server", () => {
     const toolPlanOutput = outputItem(await response.json());
     expect(toolPlanOutput.type).toBe("function_call");
     expect(toolPlanOutput.name).toBe("web_search");
-    expect(String(toolPlanOutput.arguments)).toContain("denied-input");
+    expect(String(toolPlanOutput.arguments)).toContain("OPENCLAW_QA_WEB_SEARCH_DENIED_INPUT");
   });
 
   it("plans QA subagent handoff calls even when Codex dynamic tools are not in body.tools", async () => {

extensions/qa-lab/src/providers/mock-openai/server.ts

@@ -5,6 +5,7 @@ import { setTimeout as sleep } from "node:timers/promises";
 import { escapeRegExp } from "openclaw/plugin-sdk/text-utility-runtime";
 import { readRequestBodyWithLimit } from "openclaw/plugin-sdk/webhook-ingress";
 import { closeQaHttpServer } from "../../bus-server.js";
+import { QA_LAB_WEB_SEARCH_DENIED_INPUT_QUERY } from "../../qa-web-search-provider.js";
 import { writeJson } from "../shared/http-json.js";
 
 type ResponsesInputItem = Record<string, unknown>;
@@ -860,6 +861,9 @@ function extractToolSearchTarget(text: string): string | null {
 }
 
 function buildQaToolSearchArgs(targetTool: string, failureMode: boolean): Record<string, unknown> {
+  if (failureMode && targetTool === "web_search") {
+    return { query: QA_LAB_WEB_SEARCH_DENIED_INPUT_QUERY };
+  }
   if (failureMode) {
     return { __qaFailureMode: "denied-input" };
   }
@@ -1535,49 +1539,57 @@ function buildToolCallEvents(prompt: string): StreamEvent[] {
 function buildReleaseAuditJson() {
   return `${JSON.stringify(
     {
-      verified: true,
+      verified: false,
       findings: [
         {
           id: "REL-GATEWAY-417",
           source: "src/gateway/reconnect.ts",
           status: "retry jitter verified, resume token fallback still needs manual spot check",
+          verified: true,
         },
         {
           id: "REL-CHANNEL-238",
           source: "src/channels/delivery.ts",
           status: "thread replies preserve ordering, root-channel fallback needs handoff note",
+          verified: true,
         },
         {
           id: "REL-CRON-904",
           source: "src/scheduling/cron.ts",
           status: "single-run lock verified for restart wakeups",
+          verified: true,
         },
         {
           id: "REL-MEMORY-552",
           source: "src/memory/recall.ts",
           status:
             "fallback summary survives empty memory search; ranking sample needs second reviewer",
+          verified: true,
         },
         {
           id: "REL-PLUGIN-319",
           source: "src/plugins/runtime.ts",
           status: "bundled runtime manifest loads cleanly after restart",
+          verified: true,
         },
         {
           id: "REL-INSTALL-846",
           source: "install/update.ts",
           status: "update smoke passed from previous stable tag",
+          verified: true,
         },
         {
           id: "REL-DOCS-611",
           source: "docs/operator-notes.md",
           status:
             "docs mention reconnect, cron, memory, plugin, and installer checks; channel ordering and UI notes need maintainer handoff",
+          verified: true,
         },
         {
           id: "REL-UI-BLOCKED",
           source: "ui/control-panel.ts",
           status: "blocked: source file was referenced by checklist but missing from the fixture",
+          verified: false,
         },
       ],
     },

extensions/qa-lab/src/qa-web-search-provider.test.ts

@@ -3,6 +3,7 @@ import { describe, expect, it } from "vitest";
 import { createQaLabWebSearchProvider as createQaLabWebSearchContractProvider } from "../web-search-contract-api.js";
 import {
   createQaLabWebSearchProvider,
+  QA_LAB_WEB_SEARCH_DENIED_INPUT_QUERY,
   QA_LAB_WEB_SEARCH_PROVIDER_ID,
 } from "./qa-web-search-provider.js";
 
@@ -55,4 +56,16 @@ describe("qa-lab web search provider", () => {
 
     await expect(tool.execute({ __qaFailureMode: "denied-input" })).rejects.toThrow(/query/i);
   });
+
+  it("keeps the QA failure sentinel as a deterministic tool failure", async () => {
+    const provider = createQaLabWebSearchProvider();
+    const tool = provider.createTool({});
+    if (!tool) {
+      throw new Error("expected QA Lab web search tool");
+    }
+
+    await expect(tool.execute({ query: QA_LAB_WEB_SEARCH_DENIED_INPUT_QUERY })).rejects.toThrow(
+      /denied input sentinel/i,
+    );
+  });
 });

extensions/qa-lab/src/qa-web-search-provider.ts

@@ -9,6 +9,7 @@ import {
 } from "openclaw/plugin-sdk/provider-web-search";
 
 export const QA_LAB_WEB_SEARCH_PROVIDER_ID = "qa-lab-search";
+export const QA_LAB_WEB_SEARCH_DENIED_INPUT_QUERY = "OPENCLAW_QA_WEB_SEARCH_DENIED_INPUT";
 
 const QaLabWebSearchSchema = {
   type: "object",
@@ -64,6 +65,9 @@ export function createQaLabWebSearchProvider(): WebSearchProviderPlugin {
       parameters: QaLabWebSearchSchema,
       execute: async (args) => {
         const query = readStringParam(args, "query", { required: true });
+        if (query === QA_LAB_WEB_SEARCH_DENIED_INPUT_QUERY) {
+          throw new Error("QA Lab web_search denied input sentinel");
+        }
         const count =
           readPositiveIntegerParam(args, "count", {
             max: MAX_SEARCH_COUNT,

extensions/qa-lab/src/run-config.test.ts

@@ -161,6 +161,22 @@ describe("qa run config", () => {
     expect(outputDir.startsWith(path.join(repoRoot, ".artifacts", "qa-e2e", "lab-"))).toBe(true);
   });
 
+  it("keeps generated run output dirs unique within the same millisecond", () => {
+    vi.useFakeTimers();
+    try {
+      vi.setSystemTime(new Date("2026-06-23T07:30:00.000Z"));
+      const repoRoot = path.resolve("/tmp/openclaw-repo");
+      const first = createQaRunOutputDir(repoRoot);
+      const second = createQaRunOutputDir(repoRoot);
+
+      expect(first).not.toBe(second);
+      expect(path.basename(first)).toMatch(/^lab-2026-06-23-073000000Z-[0-9a-f]{8}$/u);
+      expect(path.basename(second)).toMatch(/^lab-2026-06-23-073000000Z-[0-9a-f]{8}$/u);
+    } finally {
+      vi.useRealTimers();
+    }
+  });
+
   it("prefers the Codex OAuth default when the runtime resolver says it is available", () => {
     defaultQaRuntimeModelForMode.mockImplementation((mode, options) =>
       mode === "live-frontier"

extensions/qa-lab/src/run-config.ts

@@ -1,4 +1,5 @@
 // Qa Lab helper module supports run config behavior.
+import { randomUUID } from "node:crypto";
 import path from "node:path";
 import { uniqueStrings } from "openclaw/plugin-sdk/string-coerce-runtime";
 import { defaultQaModelForMode as defaultStaticQaModelForMode } from "./model-selection.js";
@@ -132,5 +133,5 @@ export function createIdleQaRunnerSnapshot(scenarios: QaSeedScenario[]): QaLabRu
 
 export function createQaRunOutputDir(baseDir = process.cwd()) {
   const stamp = new Date().toISOString().replaceAll(":", "").replaceAll(".", "").replace("T", "-");
-  return path.join(baseDir, ".artifacts", "qa-e2e", `lab-${stamp}`);
+  return path.join(baseDir, ".artifacts", "qa-e2e", `lab-${stamp}-${randomUUID().slice(0, 8)}`);
 }

extensions/qa-lab/src/scorecard-taxonomy.ts

@@ -10,7 +10,7 @@ export const QA_MATURITY_TAXONOMY_PATH = "taxonomy.yaml";
 export const QA_MATURITY_SCORES_PATH = "qa/maturity-scores.yaml";
 export const QA_MATURITY_SCORE_KEYS = ["quality", "completeness"] as const;
 export const QA_MATURITY_SCORE_LABELS = [
-  "Lovable",
+  "Clawesome",
   "Stable",
   "Beta",
   "Alpha",

extensions/qa-lab/src/self-check.test.ts

@@ -53,10 +53,13 @@ describe("resolveQaSelfCheckOutputPath", () => {
     ).toBe("/tmp/custom/self-check.md");
   });
 
-  it("anchors default self-check reports under the provided repo root", () => {
+  it("anchors default self-check reports under unique files in the provided repo root", () => {
     const repoRoot = path.resolve("/tmp/openclaw-repo");
-    expect(resolveQaSelfCheckOutputPath({ repoRoot })).toBe(
-      path.join(repoRoot, ".artifacts", "qa-e2e", "self-check.md"),
-    );
+    const firstPath = resolveQaSelfCheckOutputPath({ repoRoot });
+    const secondPath = resolveQaSelfCheckOutputPath({ repoRoot });
+
+    expect(path.dirname(firstPath)).toBe(path.join(repoRoot, ".artifacts", "qa-e2e"));
+    expect(path.basename(firstPath)).toMatch(/^self-check-[a-z0-9]+-[a-f0-9]{8}\.md$/u);
+    expect(secondPath).not.toBe(firstPath);
   });
 });

extensions/qa-lab/src/self-check.ts

@@ -3,6 +3,7 @@ import fs from "node:fs/promises";
 import path from "node:path";
 import type { OpenClawConfig } from "openclaw/plugin-sdk/config-contracts";
 import { renderQaMarkdownReport } from "openclaw/plugin-sdk/qa-runtime";
+import { createQaArtifactRunId } from "./artifact-run-id.js";
 import type { QaBusState } from "./bus-state.js";
 import { createQaTransportAdapter, type QaTransportId } from "./qa-transport-registry.js";
 import { runQaScenario, type QaScenarioResult } from "./scenario.js";
@@ -27,7 +28,7 @@ export function resolveQaSelfCheckOutputPath(params?: { outputPath?: string; rep
     return params.outputPath;
   }
   const repoRoot = path.resolve(params?.repoRoot ?? process.cwd());
-  return path.join(repoRoot, ".artifacts", "qa-e2e", "self-check.md");
+  return path.join(repoRoot, ".artifacts", "qa-e2e", `self-check-${createQaArtifactRunId()}.md`);
 }
 
 export async function runQaSelfCheckAgainstState(params: {

extensions/qa-lab/src/suite-launch.runtime.test.ts

@@ -28,23 +28,19 @@ async function makeTempRepo(prefix: string) {
   return repoRoot;
 }
 
-async function writeEvidence(pathLocal: string) {
-  await fs.mkdir(path.dirname(pathLocal), { recursive: true });
-  await fs.writeFile(
-    pathLocal,
-    `${JSON.stringify(
-      {
-        kind: "openclaw.qa.evidence-summary",
-        schemaVersion: 2,
-        generatedAt: "2026-06-14T00:00:00.000Z",
-        evidenceMode: "full",
-        entries: [],
-      },
-      null,
-      2,
-    )}\n`,
-    "utf8",
-  );
+async function writeEvidence(pathLocal: string, writeFile = true) {
+  const evidence = {
+    kind: "openclaw.qa.evidence-summary",
+    schemaVersion: 2,
+    generatedAt: "2026-06-14T00:00:00.000Z",
+    evidenceMode: "full",
+    entries: [],
+  };
+  if (writeFile) {
+    await fs.mkdir(path.dirname(pathLocal), { recursive: true });
+    await fs.writeFile(pathLocal, `${JSON.stringify(evidence, null, 2)}\n`, "utf8");
+  }
+  return evidence;
 }
 
 describe("qa suite runtime launcher", () => {
@@ -52,12 +48,17 @@ describe("qa suite runtime launcher", () => {
     runQaFlowSuite.mockReset();
     runQaTestFileScenarios.mockReset();
     runQaFlowSuite.mockImplementation(
-      async (params: { outputDir?: string; scenarioIds?: string[] } | undefined) => {
+      async (
+        params:
+          | { outputDir?: string; scenarioIds?: string[]; writeEvidenceFile?: boolean }
+          | undefined,
+      ) => {
         const outputDir = params?.outputDir ?? "/tmp/qa-flow";
         const evidencePath = path.join(outputDir, "qa-evidence.json");
-        await writeEvidence(evidencePath);
+        const evidence = await writeEvidence(evidencePath, params?.writeEvidenceFile);
         const scenarioIds = params?.scenarioIds ?? ["channel-chat-baseline"];
         return {
+          evidence,
           outputDir,
           evidencePath,
           reportPath: path.join(outputDir, "qa-suite-report.md"),
@@ -76,14 +77,16 @@ describe("qa suite runtime launcher", () => {
       async (params: {
         outputDir: string;
         scenarios: Array<{ id: string; execution: { kind: "script" | "vitest" | "playwright" } }>;
+        writeEvidenceFile?: boolean;
       }) => {
         const [scenario] = params.scenarios;
         if (!scenario) {
           throw new Error("expected scenario");
         }
         const evidencePath = path.join(params.outputDir, "qa-evidence.json");
-        await writeEvidence(evidencePath);
+        const evidence = await writeEvidence(evidencePath, params.writeEvidenceFile);
         return {
+          evidence,
           outputDir: params.outputDir,
           executionKind: scenario.execution.kind,
           evidencePath,
@@ -247,15 +250,27 @@ describe("qa suite runtime launcher", () => {
       expect.objectContaining({
         outputDir: path.join(outputDir, "flow"),
         scenarioIds: ["channel-chat-baseline"],
+        writeEvidenceFile: false,
       }),
     );
     expect(runQaTestFileScenarios).toHaveBeenCalledWith(
       expect.objectContaining({
         outputDir: path.join(outputDir, "playwright"),
+        writeEvidenceFile: false,
       }),
     );
     await expect(fs.access(path.join(outputDir, "qa-suite-summary.json"))).resolves.toBeUndefined();
     await expect(fs.access(path.join(outputDir, "qa-evidence.json"))).resolves.toBeUndefined();
+    await expect(fs.access(path.join(outputDir, "flow", "qa-evidence.json"))).rejects.toMatchObject(
+      {
+        code: "ENOENT",
+      },
+    );
+    await expect(
+      fs.access(path.join(outputDir, "playwright", "qa-evidence.json")),
+    ).rejects.toMatchObject({
+      code: "ENOENT",
+    });
     const summary = JSON.parse(
       await fs.readFile(path.join(outputDir, "qa-suite-summary.json"), "utf8"),
     ) as {

extensions/qa-lab/src/suite-launch.runtime.ts

@@ -175,6 +175,7 @@ async function runQaTestFileSuiteFromRuntime(params: {
     providerMode,
     primaryModel,
     scenarios: params.scenarios,
+    writeEvidenceFile: runParams?.writeEvidenceFile,
   });
 }
 
@@ -292,6 +293,13 @@ async function readQaSuiteEvidenceSummary(evidencePath: string) {
   return validateQaEvidenceSummaryJson(JSON.parse(await fs.readFile(evidencePath, "utf8")));
 }
 
+async function resolveQaSuiteResultEvidenceSummary(result: {
+  evidence?: QaEvidenceSummaryJson;
+  evidencePath: string;
+}) {
+  return result.evidence ?? (await readQaSuiteEvidenceSummary(result.evidencePath));
+}
+
 function mergeQaEvidenceSummaries(params: {
   evidenceSummaries: readonly QaEvidenceSummaryJson[];
   generatedAt: string;
@@ -489,6 +497,7 @@ async function runUnifiedQaSuite(params: {
               flowPartitions.length === 1
                 ? suitePartitionOutputDir(outputDir, "flow")
                 : flowSuitePartitionOutputDir(outputDir, partition.kind),
+            writeEvidenceFile: false,
             providerMode,
             primaryModel,
             alternateModel,
@@ -512,7 +521,7 @@ async function runUnifiedQaSuite(params: {
             }
           }
           return {
-            evidenceSummaries: [await readQaSuiteEvidenceSummary(result.evidencePath)],
+            evidenceSummaries: [await resolveQaSuiteResultEvidenceSummary(result)],
             scenarioResults,
           };
         },
@@ -530,13 +539,14 @@ async function runUnifiedQaSuite(params: {
             runParams: {
               ...params.runParams,
               outputDir: suitePartitionOutputDir(outputDir, kind),
+              writeEvidenceFile: false,
               providerMode,
               primaryModel,
               scenarioIds: testFileScenarios.map((scenario) => scenario.id),
             },
             scenarios: testFileScenarios,
           });
-          testFileEvidenceSummaries.push(await readQaSuiteEvidenceSummary(result.evidencePath));
+          testFileEvidenceSummaries.push(await resolveQaSuiteResultEvidenceSummary(result));
           testFileScenarioResults.push(
             ...result.results.map((scenarioResult) => ({
               scenarioId: scenarioResult.scenario.id,

extensions/qa-lab/src/suite-planning.test.ts

@@ -86,6 +86,22 @@ describe("qa suite planning helpers", () => {
     }
   });
 
+  it("creates unique default suite output dirs inside the repo root", async () => {
+    const repoRoot = await mkdtemp(path.join(os.tmpdir(), "qa-suite-default-root-"));
+    try {
+      const firstDir = await resolveQaSuiteOutputDir(repoRoot);
+      const secondDir = await resolveQaSuiteOutputDir(repoRoot);
+
+      expect(path.dirname(firstDir)).toBe(path.join(repoRoot, ".artifacts", "qa-e2e"));
+      expect(path.basename(firstDir)).toMatch(/^suite-[a-z0-9]+-[a-f0-9]{8}$/u);
+      expect(secondDir).not.toBe(firstDir);
+      await expect(lstat(firstDir).then((stats) => stats.isDirectory())).resolves.toBe(true);
+      await expect(lstat(secondDir).then((stats) => stats.isDirectory())).resolves.toBe(true);
+    } finally {
+      await rm(repoRoot, { recursive: true, force: true });
+    }
+  });
+
   it("rejects symlinked suite output dirs that escape the repo root", async () => {
     const repoRoot = await mkdtemp(path.join(os.tmpdir(), "qa-suite-root-"));
     const outsideRoot = await mkdtemp(path.join(os.tmpdir(), "qa-suite-outside-"));