@xn-intenton-z2a/agentic-lib 7.4.23 → 7.4.25

package/.github/agents/agent-issue-resolution.md

@@ -85,6 +85,8 @@ When writing both tests and implementation:
 **Both unit tests AND behaviour tests must pass.** The project runs `npm test` (unit tests) and
 `npm run test:behaviour` (Playwright). Both are gated — your changes must pass both.
 
+**Code coverage:** Aim for the coverage thresholds stated in the Constraints section of the prompt. Write tests that exercise the code paths you're adding or modifying.
+
 ### Test philosophy
 
 - **Unit tests** bind to the detail: exact return values, error types, edge cases, parameter validation.

package/.github/agents/agent-supervisor.md

@@ -68,7 +68,7 @@ If an **Implementation Review** section is present in the prompt, examine it car
 
 ## Priority Order
 
-1. **Always strive to close gaps** — every action you take should aim to satisfy the remaining NOT MET metrics. If the code is already complete (see Source Exports and Recently Closed Issues), use `nop` and let the director evaluate. Otherwise, create one comprehensive issue that targets the entire mission (all acceptance criteria, tests, website, docs, README). Only create a second issue if the first transform couldn't complete everything, and scope it to the remaining work. Do not create issues just to fill a quota.
+1. **Always strive to close gaps** — every action you take should aim to satisfy the remaining NOT MET metrics. If the code is already complete (see Source Exports and Recently Closed Issues), use `nop` and let the director evaluate. Otherwise, assess the full gap between current state and mission, then create as many distinct issues as needed to cover the entire gap. Ideally one comprehensive issue covering the whole gap, but if the work is naturally separable (e.g. different features, different layers), create multiple focused issues. Create up to the WIP limit. Each issue should be self-contained and independently deliverable.
 2. **Dispatch transform when ready issues exist** — transform is where code gets written. Always prefer it over maintain when there are open issues with the `ready` label.
 3. **Dispatch review after transform** — when recent workflow runs show a transform completion, dispatch review to close resolved issues and add `ready` labels to new issues. This keeps the pipeline flowing.
 4. **Fix failing PRs** — dispatch fix-code for any PR with failing checks (include pr-number).
@@ -78,7 +78,7 @@ If an **Implementation Review** section is present in the prompt, examine it car
 
 1. **Check what's already in progress** — don't duplicate work. If the workflow is already running, don't dispatch another.
 2. **Prioritise code generation** — the goal is working code. Prefer actions that produce code (dev-only, fix) over metadata (maintain, label).
-3. **Right-size the work** — break the mission into chunks just big enough to reliably deliver. One comprehensive issue is better than many small ones. Only create a follow-up issue when the previous transform has landed and gaps remain.
+3. **Right-size the work** — break the mission into the fewest chunks that can each be reliably delivered in a single transform. Create all the issues needed upfront rather than waiting for each to land before creating the next. Each issue should request maximum implementation in its scope.
 4. **Respect limits** — don't create issues beyond the WIP limit shown in the context. Don't dispatch workflows that will fail due to missing prerequisites.
 
 ## When to use each action

package/.github/workflows/agentic-lib-flow.yml

@@ -64,7 +64,7 @@ on:
         default: "max"
         options:
           - min
-          - recommended
+          - med
           - max
       workflow-runs:
         description: "Number of workflow iterations (1-16)"

package/.github/workflows/agentic-lib-init.yml

@@ -124,7 +124,7 @@ on:
         options:
           - ""
           - min
-          - recommended
+          - med
           - max
       create-seed-issues:
         description: "Create initial seed issues after purge"

package/.github/workflows/agentic-lib-schedule.yml

@@ -67,7 +67,7 @@ on:
         options:
           - ""
           - min
-          - recommended
+          - med
           - max
       dry-run:
         description: "Skip commit and push"

package/.github/workflows/agentic-lib-workflow.yml

@@ -68,7 +68,7 @@ on:
         options:
           - ""
           - min
-          - recommended
+          - med
           - max
       mode:
         description: "Run mode"
@@ -140,7 +140,10 @@ jobs:
       - uses: actions/checkout@v6
         with:
           ref: ${{ inputs.ref || github.sha }}
-          sparse-checkout: ${{ env.configPath }}
+          sparse-checkout: |
+            ${{ env.configPath }}
+            MISSION_COMPLETE.md
+            MISSION_FAILED.md
           sparse-checkout-cone-mode: false
       - name: Normalise params
         id: normalise
@@ -190,6 +193,23 @@ jobs:
           echo "log-prefix=${LOG_PREFIX:-agent-log-}" >> $GITHUB_OUTPUT
           echo "log-branch=${LOG_BRANCH:-agentic-lib-logs}" >> $GITHUB_OUTPUT
           echo "screenshot-file=${SCREENSHOT:-SCREENSHOT_INDEX.png}" >> $GITHUB_OUTPUT
+      - name: Check mission-complete signal (W4)
+        id: mission-check
+        shell: bash
+        run: |
+          SUPERVISOR=""
+          if [ -f "${{ env.configPath }}" ]; then
+            SUPERVISOR=$(grep '^\s*supervisor\s*=' "${{ env.configPath }}" 2>/dev/null | head -1 | sed 's/.*=\s*"\([^"]*\)".*/\1/' || true)
+          fi
+          if [ -f MISSION_COMPLETE.md ] && [ "$SUPERVISOR" != "maintenance" ]; then
+            echo "mission-complete=true" >> $GITHUB_OUTPUT
+            echo "::notice::Mission is complete — most jobs will be skipped"
+          elif [ -f MISSION_FAILED.md ]; then
+            echo "mission-complete=true" >> $GITHUB_OUTPUT
+            echo "::notice::Mission has failed — most jobs will be skipped"
+          else
+            echo "mission-complete=false" >> $GITHUB_OUTPUT
+          fi
     outputs:
       model: ${{ steps.normalise.outputs.model }}
       profile: ${{ steps.normalise.outputs.profile }}
@@ -203,6 +223,7 @@ jobs:
       log-prefix: ${{ steps.normalise.outputs.log-prefix }}
       log-branch: ${{ steps.normalise.outputs.log-branch }}
       screenshot-file: ${{ steps.normalise.outputs.screenshot-file }}
+      mission-complete: ${{ steps.mission-check.outputs.mission-complete }}
 
   # ─── PR Cleanup: merge/close/delete stale PRs and branches ─────────
   pr-cleanup:
@@ -280,7 +301,8 @@ jobs:
   behaviour-telemetry:
     needs: params
     if: |
-      needs.params.outputs.mode == 'full' || needs.params.outputs.mode == 'dev-only' || needs.params.outputs.mode == 'review-only'
+      needs.params.outputs.mission-complete != 'true' &&
+      (needs.params.outputs.mode == 'full' || needs.params.outputs.mode == 'dev-only' || needs.params.outputs.mode == 'review-only')
     runs-on: ubuntu-latest
     container: mcr.microsoft.com/playwright:v1.58.2-noble
     steps:
@@ -323,6 +345,7 @@ jobs:
     needs: [params, behaviour-telemetry]
     if: |
       !cancelled() &&
+      needs.params.outputs.mission-complete != 'true' &&
       (needs.params.outputs.mode == 'full' || needs.params.outputs.mode == 'dev-only' || needs.params.outputs.mode == 'review-only')
     runs-on: ubuntu-latest
     steps:
@@ -550,14 +573,25 @@ jobs:
               if (fs.existsSync('agentic-lib.toml')) {
                 const toml = fs.readFileSync('agentic-lib.toml', 'utf8');
                 const profileMatch = toml.match(/^\s*profile\s*=\s*"(\w+)"/m);
-                const profile = profileMatch ? profileMatch[1] : 'recommended';
-                const PROFILE_LIMITS = { min: 10000, recommended: 30000, max: 60000 };
+                const profile = profileMatch ? profileMatch[1] : 'med';
+                const PROFILE_LIMITS = { min: 10000, med: 30000, max: 60000 };
                 maxTelemetryChars = PROFILE_LIMITS[profile] || 30000;
               }
             } catch (e) {}
             const summary = JSON.stringify(telemetry);
             core.setOutput('telemetry', summary.slice(0, maxTelemetryChars));
 
+            // W15: Output counts for downstream job gating
+            core.setOutput('open-issue-count', String(issuesSummary.length));
+            core.setOutput('open-pr-count', String(prsSummary.length));
+
+            // W19: Output unit test summary for transform prompt
+            const unitSummary = `exit=${telemetry.liveTests.unit.exitCode} pass=${telemetry.liveTests.unit.passCount} fail=${telemetry.liveTests.unit.failCount}`;
+            core.setOutput('unit-test-summary', unitSummary);
+            // Truncated unit test output for transform (first 4000 chars)
+            const unitOutputForTransform = (telemetry.liveTests?.unit?.output || '').substring(0, 4000);
+            core.setOutput('unit-test-output', unitOutputForTransform);
+
       - name: Output telemetry summary
         shell: bash
         run: |
@@ -567,6 +601,10 @@ jobs:
 
     outputs:
       telemetry: ${{ steps.gather.outputs.telemetry }}
+      open-issue-count: ${{ steps.gather.outputs.open-issue-count }}
+      open-pr-count: ${{ steps.gather.outputs.open-pr-count }}
+      unit-test-summary: ${{ steps.gather.outputs.unit-test-summary }}
+      unit-test-output: ${{ steps.gather.outputs.unit-test-output }}
 
   # ─── Maintain: features + library (push to main) ───────────────────
   # Runs early (parallel with pr-cleanup/telemetry) so supervisor sees features.
@@ -574,6 +612,7 @@ jobs:
     needs: [params]
     if: |
       !cancelled() &&
+      needs.params.outputs.mission-complete != 'true' &&
       (needs.params.outputs.mode == 'full' || needs.params.outputs.mode == 'maintain-only') &&
       needs.params.result == 'success' &&
       inputs.skipMaintain != 'true' && inputs.skipMaintain != true
@@ -656,6 +695,7 @@ jobs:
       - name: Maintain library
         id: maintain-library
         if: steps.mission-check.outputs.mission-complete != 'true'
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -670,6 +710,7 @@ jobs:
       - name: Maintain features
         id: maintain-features
         if: steps.mission-check.outputs.mission-complete != 'true'
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -710,6 +751,7 @@ jobs:
     needs: [params]
     if: |
       !cancelled() &&
+      needs.params.outputs.mission-complete != 'true' &&
       (needs.params.outputs.mode == 'full' || needs.params.outputs.mode == 'maintain-only') &&
       needs.params.result == 'success'
     runs-on: ubuntu-latest
@@ -755,6 +797,7 @@ jobs:
       - name: Run implementation review
         id: review
         if: github.repository != 'xn-intenton-z2a/agentic-lib'
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -776,6 +819,7 @@ jobs:
     needs: [params, telemetry, maintain, implementation-review]
     if: |
       !cancelled() &&
+      needs.params.outputs.mission-complete != 'true' &&
       (needs.params.outputs.mode == 'full' || needs.params.outputs.mode == 'dev-only') &&
       needs.params.result == 'success'
     runs-on: ubuntu-latest
@@ -816,6 +860,7 @@ jobs:
       - name: Run director
         id: director
         if: github.repository != 'xn-intenton-z2a/agentic-lib'
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -903,6 +948,7 @@ jobs:
 
       - name: Run supervisor
         if: github.repository != 'xn-intenton-z2a/agentic-lib'
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -1179,6 +1225,7 @@ jobs:
           env.FIX_PR_NUMBER != '' &&
           steps.fix-mission-check.outputs.mission-complete != 'true' &&
           env.FIX_REASON == 'requested'
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -1196,6 +1243,7 @@ jobs:
         if: |
           env.FIX_MAIN_BUILD == 'true' &&
           steps.fix-mission-check.outputs.mission-complete != 'true'
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -1229,6 +1277,63 @@ jobs:
             echo "No additional changes to push"
           fi
 
+      # W20: Immediately attempt merge after fix-stuck resolves conflicts
+      - name: "W20: Immediate merge attempt after fix"
+        if: |
+          github.repository != 'xn-intenton-z2a/agentic-lib' &&
+          env.FIX_PR_NUMBER != '' &&
+          steps.fix-mission-check.outputs.mission-complete != 'true' &&
+          needs.params.outputs.dry-run != 'true'
+        uses: actions/github-script@v8
+        with:
+          script: |
+            const owner = context.repo.owner;
+            const repo = context.repo.repo;
+            const prNumber = parseInt('${{ env.FIX_PR_NUMBER }}');
+            if (!prNumber) return;
+
+            // Wait for checks to register
+            await new Promise(r => setTimeout(r, 15000));
+
+            for (let attempt = 0; attempt < 3; attempt++) {
+              const { data: pr } = await github.rest.pulls.get({
+                owner, repo, pull_number: prNumber,
+              });
+
+              if (pr.mergeable && pr.mergeable_state === 'clean') {
+                try {
+                  await github.rest.pulls.merge({
+                    owner, repo, pull_number: prNumber, merge_method: 'squash',
+                  });
+                  core.info(`W20: Merged PR #${prNumber} immediately after fix`);
+                  try {
+                    await github.rest.git.deleteRef({ owner, repo, ref: `heads/${pr.head.ref}` });
+                  } catch (e) { /* branch may already be deleted */ }
+                  // Label associated issue
+                  const branchPrefix = 'agentic-lib-issue-';
+                  if (pr.head.ref.startsWith(branchPrefix)) {
+                    const issueNum = parseInt(pr.head.ref.replace(branchPrefix, ''));
+                    if (issueNum) {
+                      try {
+                        await github.rest.issues.addLabels({ owner, repo, issue_number: issueNum, labels: ['merged'] });
+                        await github.rest.issues.removeLabel({ owner, repo, issue_number: issueNum, name: 'in-progress' });
+                      } catch (e) { /* label may not exist */ }
+                    }
+                  }
+                  return;
+                } catch (e) {
+                  core.info(`W20: Merge attempt ${attempt + 1} failed: ${e.message}`);
+                }
+              } else if (pr.mergeable_state === 'unstable' || pr.mergeable === null) {
+                core.info(`W20: PR not ready yet (${pr.mergeable_state}), waiting...`);
+                await new Promise(r => setTimeout(r, 15000));
+              } else {
+                core.info(`W20: PR not mergeable (${pr.mergeable_state}), leaving for pr-cleanup`);
+                break;
+              }
+            }
+            core.info(`W20: PR #${prNumber} left open for next cycle`);
+
       - name: Commit, push, and open PR for main build fix
         if: github.repository != 'xn-intenton-z2a/agentic-lib' && env.FIX_MAIN_BUILD == 'true' && steps.fix-mission-check.outputs.mission-complete != 'true'
         env:
@@ -1249,12 +1354,14 @@ jobs:
             --label automerge
 
   # ─── Review: close resolved issues, enhance with criteria ──────────
+  # W15: Skip review when there are no open issues to review
   review-features:
-    needs: [params, maintain, supervisor]
+    needs: [params, maintain, supervisor, telemetry]
     if: |
       !cancelled() &&
       (needs.params.outputs.mode == 'full' || needs.params.outputs.mode == 'review-only') &&
-      needs.params.result == 'success'
+      needs.params.result == 'success' &&
+      needs.telemetry.outputs.open-issue-count != '0'
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v6
@@ -1278,6 +1385,7 @@ jobs:
           fi
 
       - name: Review issues
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -1289,6 +1397,7 @@ jobs:
           model: ${{ needs.params.outputs.model }}
 
       - name: Enhance issues
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
@@ -1299,9 +1408,9 @@ jobs:
           instructions: ".github/agents/agent-ready-issue.md"
           model: ${{ needs.params.outputs.model }}
 
-  # ─── Dev: sequential issue resolution ──────────────────────────────
+  # ─── Dev: issue resolution (W7: multiple issues per session) ──────
   dev:
-    needs: [params, maintain, review-features]
+    needs: [params, maintain, review-features, telemetry, implementation-review]
     if: |
       !cancelled() &&
       (needs.params.outputs.mode == 'full' || needs.params.outputs.mode == 'dev-only') &&
@@ -1383,25 +1492,32 @@ jobs:
               core.setOutput('issue-number', specificIssue);
               return;
             }
-            // W7: Mechanical instability override — prioritise instability issues
-            // before any other ready issues, regardless of supervisor decisions
+            // W7: Collect ALL ready issues for concurrent resolution in one session
+            const collected = [];
+            // Instability issues first (mechanical priority override)
             const { data: instabilityIssues } = await github.rest.issues.listForRepo({
               ...context.repo, state: 'open', labels: 'instability',
-              sort: 'created', direction: 'asc', per_page: 1,
+              sort: 'created', direction: 'asc', per_page: 10,
             });
-            if (instabilityIssues.length > 0) {
-              core.setOutput('issue-number', String(instabilityIssues[0].number));
-              core.info(`Instability override: targeting issue #${instabilityIssues[0].number}: ${instabilityIssues[0].title}`);
-              return;
+            for (const i of instabilityIssues) {
+              collected.push(i.number);
+              core.info(`Instability issue: #${i.number}: ${i.title}`);
             }
-            // Find oldest open issue with 'ready' label
-            const { data: issues } = await github.rest.issues.listForRepo({
+            // Then ready issues
+            const { data: readyIssues } = await github.rest.issues.listForRepo({
               ...context.repo, state: 'open', labels: 'ready',
-              sort: 'created', direction: 'asc', per_page: 1,
+              sort: 'created', direction: 'asc', per_page: 10,
             });
-            if (issues.length > 0) {
-              core.setOutput('issue-number', String(issues[0].number));
-              core.info(`Targeting issue #${issues[0].number}: ${issues[0].title}`);
+            for (const i of readyIssues) {
+              if (!collected.includes(i.number)) {
+                collected.push(i.number);
+                core.info(`Ready issue: #${i.number}: ${i.title}`);
+              }
+            }
+            if (collected.length > 0) {
+              // W7: Pass all issues as comma-separated list
+              core.setOutput('issue-number', collected.join(','));
+              core.info(`Targeting ${collected.length} issue(s): ${collected.join(', ')}`);
             } else {
               core.setOutput('issue-number', '');
               core.info('No ready issues found');
@@ -1411,17 +1527,24 @@ jobs:
         if: steps.issue.outputs.issue-number != ''
         id: branch
         run: |
-          ISSUE_NUMBER="${{ steps.issue.outputs.issue-number }}"
+          # W7: Use first issue number for branch name (may be comma-separated list)
+          ISSUE_LIST="${{ steps.issue.outputs.issue-number }}"
+          ISSUE_NUMBER="${ISSUE_LIST%%,*}"
           BRANCH="agentic-lib-issue-${ISSUE_NUMBER}"
           git checkout -b "${BRANCH}" 2>/dev/null || git checkout "${BRANCH}"
           echo "branchName=${BRANCH}" >> $GITHUB_OUTPUT
 
       - name: Run transformation
         if: steps.issue.outputs.issue-number != ''
+        timeout-minutes: 10
         uses: ./.github/agentic-lib/actions/agentic-step
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           COPILOT_GITHUB_TOKEN: ${{ secrets.COPILOT_GITHUB_TOKEN }}
+          REVIEW_ADVICE: ${{ needs.implementation-review.outputs.review-advice }}
+          REVIEW_GAPS: ${{ needs.implementation-review.outputs.review-gaps }}
+          TELEMETRY_UNIT_TEST_SUMMARY: ${{ needs.telemetry.outputs.unit-test-summary }}
+          TELEMETRY_UNIT_TEST_OUTPUT: ${{ needs.telemetry.outputs.unit-test-output }}
         with:
           task: "transform"
           config: ${{ needs.params.outputs.config-path }}
@@ -1501,7 +1624,10 @@ jobs:
             const owner = context.repo.owner;
             const repo = context.repo.repo;
             const branchName = '${{ steps.branch.outputs.branchName }}';
-            const issueNumber = '${{ steps.issue.outputs.issue-number }}';
+            const issueList = '${{ steps.issue.outputs.issue-number }}';
+            // W7: issueNumber may be comma-separated list
+            const issueNumbers = issueList.split(',').map(n => n.trim()).filter(Boolean);
+            const issueNumber = issueNumbers[0] || '';
 
             if (!branchName) return;
 
@@ -1525,6 +1651,12 @@ jobs:
               head: `${owner}:${branchName}`, per_page: 1,
             });
 
+            // W7: Build PR body with Closes for all issues
+            const closesLines = issueNumbers.map(n => `Closes #${n}`).join('\n');
+            const prTitle = issueNumbers.length > 1
+              ? `fix: resolve issues ${issueNumbers.map(n => '#' + n).join(', ')}`
+              : `fix: resolve issue #${issueNumber}`;
+
             let prNumber;
             if (existingPRs.length > 0) {
               prNumber = existingPRs[0].number;
@@ -1532,8 +1664,8 @@ jobs:
             } else {
               const { data: pr } = await github.rest.pulls.create({
                 owner, repo,
-                title: `fix: resolve issue #${issueNumber}`,
-                body: `Closes #${issueNumber}\n\nAutomated transformation.`,
+                title: prTitle,
+                body: `${closesLines}\n\nAutomated transformation.`,
                 head: branchName, base: 'main',
               });
               prNumber = pr.number;
@@ -1583,7 +1715,7 @@ jobs:
 
   # ─── Post-merge: stats, schedule, mission check ────────────────────
   post-merge:
-    needs: [params, maintain, dev, pr-cleanup]
+    needs: [params, maintain, dev, pr-cleanup, implementation-review]
     if: ${{ !cancelled() && needs.params.result == 'success' }}
     runs-on: ubuntu-latest
     steps:
@@ -1605,6 +1737,75 @@ jobs:
           echo "- Dry-run: ${{ needs.params.outputs.dry-run }}" >> $GITHUB_STEP_SUMMARY
           echo "- Website: [${SITE_URL}](${SITE_URL})" >> $GITHUB_STEP_SUMMARY
 
+      # W14: Post-merge director check — re-evaluate mission status after dev/PR merges
+      - name: Fetch log and screenshot from log branch (post-merge director)
+        if: |
+          needs.params.outputs.mission-complete != 'true' &&
+          needs.params.outputs.dry-run != 'true' &&
+          github.repository != 'xn-intenton-z2a/agentic-lib'
+        env:
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+          SCREENSHOT_FILE: ${{ needs.params.outputs.screenshot-file }}
+        run: |
+          git fetch origin "${LOG_BRANCH}" 2>/dev/null || true
+          git show "origin/${LOG_BRANCH}:${SCREENSHOT_FILE}" > "${SCREENSHOT_FILE}" 2>/dev/null || true
+          git show "origin/${LOG_BRANCH}:agentic-lib-state.toml" > "agentic-lib-state.toml" 2>/dev/null || true
+
+      - uses: actions/setup-node@v6
+        if: |
+          needs.params.outputs.mission-complete != 'true' &&
+          needs.params.outputs.dry-run != 'true' &&
+          github.repository != 'xn-intenton-z2a/agentic-lib'
+        with:
+          node-version: "24"
+
+      - name: Self-init (agentic-lib dev only)
+        if: |
+          needs.params.outputs.mission-complete != 'true' &&
+          needs.params.outputs.dry-run != 'true' &&
+          hashFiles('scripts/self-init.sh') != '' && hashFiles('.github/agentic-lib/actions/agentic-step/package.json') == ''
+        run: bash scripts/self-init.sh
+
+      - name: Install agentic-step dependencies (post-merge director)
+        if: |
+          needs.params.outputs.mission-complete != 'true' &&
+          needs.params.outputs.dry-run != 'true' &&
+          github.repository != 'xn-intenton-z2a/agentic-lib'
+        working-directory: .github/agentic-lib/actions/agentic-step
+        run: |
+          npm ci
+          if [ -d "../../copilot" ]; then
+            ln -sf "$(pwd)/node_modules" ../../copilot/node_modules
+          fi
+
+      - name: "W14: Post-merge director evaluation"
+        id: post-merge-director
+        if: |
+          needs.params.outputs.mission-complete != 'true' &&
+          needs.params.outputs.dry-run != 'true' &&
+          github.repository != 'xn-intenton-z2a/agentic-lib'
+        timeout-minutes: 10
+        uses: ./.github/agentic-lib/actions/agentic-step
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          COPILOT_GITHUB_TOKEN: ${{ secrets.COPILOT_GITHUB_TOKEN }}
+          REVIEW_ADVICE: ${{ needs.implementation-review.outputs.review-advice }}
+          REVIEW_GAPS: ${{ needs.implementation-review.outputs.review-gaps }}
+        with:
+          task: "direct"
+          config: ${{ needs.params.outputs.config-path }}
+          instructions: ".github/agents/agent-director.md"
+          model: ${{ needs.params.outputs.model }}
+
+      - name: Push log to log branch (post-merge director)
+        if: |
+          needs.params.outputs.mission-complete != 'true' &&
+          needs.params.outputs.dry-run != 'true' &&
+          github.repository != 'xn-intenton-z2a/agentic-lib'
+        env:
+          LOG_BRANCH: ${{ needs.params.outputs.log-branch }}
+        run: bash .github/agentic-lib/scripts/push-to-logs.sh agent-log-*.md agentic-lib-state.toml
+
   # ─── Post-commit validation: call test workflow to verify branch health ───
   post-commit-test:
     needs: [params, maintain, dev, fix-stuck, post-merge]

package/agentic-lib.toml

@@ -36,9 +36,9 @@ test = "npm ci && npm test"
 # library-limit = 32
 
 [tuning]
-# Profile sets defaults for all tuning and limit knobs: min | recommended | max
+# Profile sets defaults for all tuning and limit knobs: min | med | max
 # Profile definitions live in [profiles.*] sections below.
-profile = "min"                                  #@dist "recommended"
+profile = "min"                                  #@dist "max"
 #
 # Model selection — each has different strengths:
 #   gpt-5-mini      — Fast, cheap, supports reasoning-effort. Best for CI and iteration.
@@ -55,6 +55,13 @@ infinite-sessions = false        # set to true for long sessions with compaction
 # max-issues = 5
 # stale-days = 14
 # max-discussion-comments = 5
+# session-timeout-ms = 480000    # LLM session timeout in ms (should be < workflow step timeout)
+# max-tokens = 200000            # token budget — controls max tool calls (tokens / 5000)
+# max-read-chars = 20000         # max chars per read_file result
+# max-test-output = 4000         # max chars of test output in prompts
+# max-file-listing = 30          # max files in directory listings (0 = unlimited)
+# max-library-index = 2000       # max chars for library index summary
+# max-fix-test-output = 8000     # max chars of failed run log in fix-code
 
 # ─── Profile Definitions ────────────────────────────────────────────
 # Each profile defines tuning and limits defaults. The active profile
@@ -75,9 +82,16 @@ max-attempts-per-branch = 2               # max transform attempts before abando
 max-attempts-per-issue = 1                 # max transform attempts before abandoning an issue
 features-limit = 2                     # max feature files in features/ directory
 library-limit = 8                      # max library entries in library/ directory
+session-timeout-ms = 480000            # LLM session timeout in ms (8 min, below 10-min workflow step)
+max-tokens = 200000                    # token budget for tool-call cap calculation
+max-read-chars = 20000                 # max chars returned from read_file tool
+max-test-output = 4000                 # max chars of test output in prompts
+max-file-listing = 30                  # max files in directory listings (0 = unlimited)
+max-library-index = 2000               # max chars for library index in prompts
+max-fix-test-output = 8000             # max chars of failed run log in fix-code
 
-[profiles.recommended]
-# Balanced — good results, default for consumer repos.
+[profiles.med]
+# Balanced — good results, middle ground.
 reasoning-effort = "medium"            # low | medium | high | none
 infinite-sessions = true               # enable session compaction for long runs
 transformation-budget = 32             # max code-changing cycles per run
@@ -90,6 +104,13 @@ max-attempts-per-branch = 3               # max transform attempts before abando
 max-attempts-per-issue = 2                 # max transform attempts before abandoning an issue
 features-limit = 4                     # max feature files in features/ directory
 library-limit = 32                     # max library entries in library/ directory
+session-timeout-ms = 480000            # LLM session timeout in ms (8 min, below 10-min workflow step)
+max-tokens = 200000                    # token budget for tool-call cap calculation
+max-read-chars = 50000                 # max chars returned from read_file tool
+max-test-output = 10000                # max chars of test output in prompts
+max-file-listing = 100                 # max files in directory listings (0 = unlimited)
+max-library-index = 5000               # max chars for library index in prompts
+max-fix-test-output = 15000            # max chars of failed run log in fix-code
 
 [profiles.max]
 # Thorough — maximum context for complex missions.
@@ -105,6 +126,18 @@ max-attempts-per-branch = 5               # max transform attempts before abando
 max-attempts-per-issue = 4                 # max transform attempts before abandoning an issue
 features-limit = 8                     # max feature files in features/ directory
 library-limit = 64                     # max library entries in library/ directory
+session-timeout-ms = 480000            # LLM session timeout in ms (8 min, below 10-min workflow step)
+max-tokens = 500000                    # token budget for tool-call cap calculation
+max-read-chars = 100000                # max chars returned from read_file tool
+max-test-output = 20000                # max chars of test output in prompts
+max-file-listing = 0                   # max files in directory listings (0 = unlimited)
+max-library-index = 10000              # max chars for library index in prompts
+max-fix-test-output = 30000            # max chars of failed run log in fix-code
+
+[goals]
+# W12/W13: Code coverage thresholds — stated in all code-changing prompts
+min-line-coverage = 50             # minimum % line coverage required
+min-branch-coverage = 30           # minimum % branch coverage required
 
 [mission-complete]
 # Thresholds for deterministic mission-complete declaration.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xn-intenton-z2a/agentic-lib",
-  "version": "7.4.23",
+  "version": "7.4.25",
   "description": "Agentic-lib Agentic Coding Systems SDK powering automated GitHub workflows.",
   "type": "module",
   "scripts": {

package/src/actions/agentic-step/tasks/direct.js

@@ -213,6 +213,45 @@ async function executeMissionComplete(octokit, repo, reason) {
   } catch (err) {
     core.warning(`Could not commit MISSION_COMPLETE.md: ${err.message}`);
   }
+
+  // W2: Update persistent state (Benchmark 011 FINDING-3)
+  try {
+    const { readState, writeState } = await import("../../../copilot/state.js");
+    const state = readState(".");
+    state.status["mission-complete"] = true;
+    state.schedule["auto-disabled"] = true;
+    state.schedule["auto-disabled-reason"] = "mission-complete";
+    writeState(".", state);
+    core.info("State updated: mission-complete, schedule auto-disabled");
+  } catch (err) {
+    core.warning(`Could not update state for mission-complete: ${err.message}`);
+  }
+
+  // W3: Disable schedule on mission-complete (Benchmark 011 FINDING-4)
+  try {
+    await octokit.rest.actions.createWorkflowDispatch({
+      ...repo,
+      workflow_id: "agentic-lib-schedule.yml",
+      ref: "main",
+      inputs: { frequency: "off" },
+    });
+    core.info("Dispatched schedule change to off after mission-complete");
+  } catch (err) {
+    core.warning(`Could not dispatch schedule change: ${err.message}`);
+  }
+
+  // W16: Notify bot about mission-complete
+  try {
+    await octokit.rest.actions.createWorkflowDispatch({
+      ...repo,
+      workflow_id: "agentic-lib-bot.yml",
+      ref: "main",
+      inputs: { message: `Mission complete: ${reason.substring(0, 200)}` },
+    });
+    core.info("Dispatched bot notification for mission-complete");
+  } catch (err) {
+    core.warning(`Could not dispatch bot notification: ${err.message}`);
+  }
 }
 
 /**
@@ -287,6 +326,19 @@ async function executeMissionFailed(octokit, repo, reason, metricAssessment) {
   } catch (err) {
     core.warning(`Could not dispatch schedule change: ${err.message}`);
   }
+
+  // W16: Notify bot about mission-failed
+  try {
+    await octokit.rest.actions.createWorkflowDispatch({
+      ...repo,
+      workflow_id: "agentic-lib-bot.yml",
+      ref: "main",
+      inputs: { message: `Mission failed: ${metricDetail.substring(0, 200)}` },
+    });
+    core.info("Dispatched bot notification for mission-failed");
+  } catch (err) {
+    core.warning(`Could not dispatch bot notification: ${err.message}`);
+  }
 }
 
 /**

package/src/actions/agentic-step/tasks/fix-code.js

@@ -6,7 +6,7 @@
 // failures, writes fixes, and runs tests via tools.
 
 import * as core from "@actions/core";
-import { readFileSync } from "fs";
+import { readFileSync, existsSync, readdirSync } from "fs";
 import { execSync } from "child_process";
 import { formatPathsSection, extractNarrative, NARRATIVE_INSTRUCTION } from "../copilot.js";
 import { runCopilotSession } from "../../../copilot/copilot-session.js";
@@ -25,14 +25,15 @@ function extractRunId(detailsUrl) {
 /**
  * Fetch actual test output from a GitHub Actions run log.
  */
-function fetchRunLog(runId) {
+// W22: maxChars configurable via profile
+function fetchRunLog(runId, maxChars = 8000) {
   try {
     const output = execSync(`gh run view ${runId} --log-failed`, {
       encoding: "utf8",
       timeout: 30000,
       env: { ...process.env },
     });
-    return output.substring(0, 8000);
+    return output.substring(0, maxChars);
   } catch (err) {
     core.debug(`[fix-code] Could not fetch log for run ${runId}: ${err.message}`);
     return null;
@@ -138,7 +139,8 @@ async function resolveConflicts({ config, pr, prNumber, instructions, model, wri
  * Fix a broken main branch build.
  */
 async function fixMainBuild({ config, runId, instructions, model, writablePaths, testCommand, octokit, repo, logFilePath, screenshotFilePath }) {
-  const logContent = fetchRunLog(runId);
+  const t = config.tuning || {};
+  const logContent = fetchRunLog(runId, t.maxFixTestOutput || 8000);
   if (!logContent) {
     core.info(`Could not fetch log for run ${runId}. Returning nop.`);
     return { outcome: "nop", details: `Could not fetch log for run ${runId}` };
@@ -169,7 +171,6 @@ async function fixMainBuild({ config, runId, instructions, model, writablePaths,
     "- Do not introduce new features — focus on making the build green",
   ].join("\n");
 
-  const t = config.tuning || {};
   const systemPrompt =
     `You are an autonomous coding agent fixing a broken build on the main branch. The test/build workflow has failed. Analyze the error log and make minimal, targeted changes to fix it.` +
     NARRATIVE_INSTRUCTION;
@@ -254,7 +255,7 @@ export async function fixCode(context) {
       const runId = extractRunId(cr.details_url);
       let logContent = null;
       if (runId) {
-        logContent = fetchRunLog(runId);
+        logContent = fetchRunLog(runId, (config.tuning || {}).maxFixTestOutput || 8000);
       }
       const detail = logContent || cr.output?.summary || "Failed";
       return `**${cr.name}:**\n${detail}`;

package/src/actions/agentic-step/tasks/maintain-features.js

@@ -16,11 +16,12 @@ import { checkWipLimit } from "../safety.js";
 /**
  * Build a file listing summary (names + sizes, not content).
  */
-function buildFileListing(dirPath, extension) {
+// W22: maxFiles configurable via profile (0 = unlimited)
+function buildFileListing(dirPath, extension, maxFiles = 30) {
   if (!dirPath || !existsSync(dirPath)) return [];
   try {
     const files = readdirSync(dirPath, { recursive: true });
-    return files
+    const filtered = files
       .filter((f) => String(f).endsWith(extension))
       .map((f) => {
         const fullPath = join(dirPath, String(f));
@@ -30,8 +31,8 @@ function buildFileListing(dirPath, extension) {
         } catch {
           return String(f);
         }
-      })
-      .slice(0, 30);
+      });
+    return maxFiles > 0 ? filtered.slice(0, maxFiles) : filtered;
   } catch {
     return [];
   }

package/src/actions/agentic-step/tasks/maintain-library.js

@@ -14,11 +14,12 @@ import { runCopilotSession } from "../../../copilot/copilot-session.js";
 /**
  * Build a file listing summary (names + sizes, not content).
  */
-function buildFileListing(dirPath, extension) {
+// W22: maxFiles configurable via profile (0 = unlimited)
+function buildFileListing(dirPath, extension, maxFiles = 30) {
   if (!dirPath || !existsSync(dirPath)) return [];
   try {
     const files = readdirSync(dirPath, { recursive: true });
-    return files
+    const filtered = files
       .filter((f) => String(f).endsWith(extension))
       .map((f) => {
         const fullPath = join(dirPath, String(f));
@@ -28,8 +29,8 @@ function buildFileListing(dirPath, extension) {
         } catch {
           return String(f);
         }
-      })
-      .slice(0, 30);
+      });
+    return maxFiles > 0 ? filtered.slice(0, maxFiles) : filtered;
   } catch {
     return [];
   }

package/src/actions/agentic-step/tasks/supervise.js

@@ -651,6 +651,31 @@ async function executeCreateIssue(octokit, repo, params, ctx) {
   }
   const body = bodyParts.join("\n");
 
+  // W5: Dedup guard against open issues — skip if a similarly-titled issue already exists
+  try {
+    const { data: openIssues } = await octokit.rest.issues.listForRepo({
+      ...repo,
+      state: "open",
+      labels: "automated",
+      sort: "created",
+      direction: "desc",
+      per_page: 20,
+    });
+    const titleLower = title.toLowerCase();
+    const titlePrefix = titleLower.substring(0, 30);
+    const openDuplicate = openIssues.find(
+      (i) =>
+        !i.pull_request &&
+        (i.title.toLowerCase().includes(titlePrefix) || titleLower.includes(i.title.toLowerCase().substring(0, 30))),
+    );
+    if (openDuplicate) {
+      core.info(`Skipping duplicate issue (similar to open #${openDuplicate.number}: "${openDuplicate.title}")`);
+      return `skipped:duplicate-open-#${openDuplicate.number}`;
+    }
+  } catch (err) {
+    core.warning(`Open issue dedup check failed: ${err.message}`);
+  }
+
   // Dedup guard: skip if a similarly-titled issue was closed in the last hour
   // Exclude issues closed before the init timestamp (cross-scenario protection)
   try {

package/src/actions/agentic-step/tasks/transform.js

@@ -8,19 +8,21 @@
 import * as core from "@actions/core";
 import { existsSync, readFileSync, readdirSync, statSync } from "fs";
 import { join, resolve } from "path";
+import { execSync } from "child_process";
 import { readOptionalFile, formatPathsSection, extractNarrative, NARRATIVE_INSTRUCTION } from "../copilot.js";
 import { runCopilotSession } from "../../../copilot/copilot-session.js";
 import { createGitHubTools, createGitTools } from "../../../copilot/github-tools.js";
 
 /**
  * Build a file listing summary (names + sizes, not content) for the lean prompt.
+ * W22: maxFiles configurable via profile (0 = unlimited).
  */
-function buildFileListing(dirPath, extensions) {
+function buildFileListing(dirPath, extensions, maxFiles = 30) {
   if (!dirPath || !existsSync(dirPath)) return [];
   const exts = Array.isArray(extensions) ? extensions : [extensions];
   try {
     const files = readdirSync(dirPath, { recursive: true });
-    return files
+    const filtered = files
       .filter((f) => exts.some((ext) => String(f).endsWith(ext)))
       .map((f) => {
         const fullPath = join(dirPath, String(f));
@@ -31,17 +33,18 @@ function buildFileListing(dirPath, extensions) {
         } catch {
           return String(f);
         }
-      })
-      .slice(0, 30); // cap listing at 30 files
+      });
+    return maxFiles > 0 ? filtered.slice(0, maxFiles) : filtered;
   } catch {
     return [];
   }
 }
 
 /**
- * Build a library index: filename + first 2 lines of each library doc, capped at 2000 chars.
+ * Build a library index: filename + first 2 lines of each library doc.
+ * W22: maxChars configurable via profile.
  */
-function buildLibraryIndex(libraryPath) {
+function buildLibraryIndex(libraryPath, maxChars = 2000) {
   if (!libraryPath || !existsSync(libraryPath)) return "";
   try {
     const files = readdirSync(libraryPath).filter((f) => f.endsWith(".md")).sort();
@@ -54,7 +57,7 @@ function buildLibraryIndex(libraryPath) {
         const content = readFileSync(fullPath, "utf8");
         const lines = content.split("\n").slice(0, 2).join(" ").trim();
         const entry = `- ${f}: ${lines}`;
-        if (totalLen + entry.length > 2000) break;
+        if (totalLen + entry.length > maxChars) break;
         entries.push(entry);
         totalLen += entry.length;
       } catch {
@@ -67,6 +70,18 @@ function buildLibraryIndex(libraryPath) {
   }
 }
 
+/**
+ * W9: Get worktree file listing via git ls-files.
+ */
+function getWorktreeFiles() {
+  try {
+    const gitFiles = execSync("git ls-files", { encoding: "utf8", timeout: 10000 }).trim();
+    return gitFiles.split("\n").filter(Boolean);
+  } catch {
+    return [];
+  }
+}
+
 /**
  * Run the full transformation pipeline from mission to code.
  *
@@ -76,6 +91,8 @@ function buildLibraryIndex(libraryPath) {
 export async function transform(context) {
   const { config, instructions, writablePaths, testCommand, model, octokit, repo, issueNumber, logFilePath, screenshotFilePath } = context;
   const t = config.tuning || {};
+  const maxFileListing = t.maxFileListing ?? 30;
+  const maxLibraryIdx = t.maxLibraryIndex || 2000;
 
   // Read mission (required)
   const mission = readOptionalFile(config.paths.mission.path);
@@ -90,23 +107,24 @@ export async function transform(context) {
     return { outcome: "nop", details: "Mission already complete (MISSION_COMPLETE.md signal)" };
   }
 
-  // Fetch target issue if specified
-  let targetIssueSection = "";
-  if (issueNumber) {
+  // W7: Fetch all target issues (supports comma-separated list)
+  const issueNumbers = issueNumber
+    ? String(issueNumber).split(",").map((n) => n.trim()).filter(Boolean)
+    : [];
+  const targetIssueSections = [];
+  for (const num of issueNumbers) {
     try {
       const { data: issue } = await octokit.rest.issues.get({
         ...repo,
-        issue_number: Number(issueNumber),
+        issue_number: Number(num),
       });
-      targetIssueSection = [
-        `## Target Issue #${issue.number}: ${issue.title}`,
+      targetIssueSections.push([
+        `### Issue #${issue.number}: ${issue.title}`,
         issue.body || "(no description)",
         `Labels: ${issue.labels.map((l) => l.name).join(", ") || "none"}`,
-        "",
-        "**Focus your transformation on resolving this specific issue.**",
-      ].join("\n");
+      ].join("\n"));
     } catch (err) {
-      core.warning(`Could not fetch target issue #${issueNumber}: ${err.message}`);
+      core.warning(`Could not fetch target issue #${num}: ${err.message}`);
     }
   }
 
@@ -114,17 +132,36 @@ export async function transform(context) {
     instructions || "Transform the repository toward its mission by identifying the next best action.";
 
   // ── Build lean prompt (structure + mission, not file contents) ──────
-  const sourceFiles = buildFileListing(config.paths.source.path, [".js", ".ts"]);
-  const testFiles = buildFileListing(config.paths.tests.path, [".js", ".ts"]);
-  const webFiles = buildFileListing(config.paths.web?.path || "src/web/", [".html", ".css", ".js"]);
-  const featureFiles = buildFileListing(config.paths.features.path, [".md"]);
-  const libraryIndex = buildLibraryIndex(config.paths.library?.path || "library/");
+  const sourceFiles = buildFileListing(config.paths.source.path, [".js", ".ts"], maxFileListing);
+  const testFiles = buildFileListing(config.paths.tests.path, [".js", ".ts"], maxFileListing);
+  const webFiles = buildFileListing(config.paths.web?.path || "src/web/", [".html", ".css", ".js"], maxFileListing);
+  const featureFiles = buildFileListing(config.paths.features.path, [".md"], maxFileListing);
+  const libraryIndex = buildLibraryIndex(config.paths.library?.path || "library/", maxLibraryIdx);
+
+  // W9: worktree file listing
+  const worktreeFiles = getWorktreeFiles();
+
+  // W17: Implementation review results from upstream
+  const reviewAdvice = process.env.REVIEW_ADVICE || "";
+  const reviewGapsRaw = process.env.REVIEW_GAPS || "";
+
+  // W19: Telemetry test output from upstream
+  const telemetryTestSummary = process.env.TELEMETRY_UNIT_TEST_SUMMARY || "";
+  const telemetryTestOutput = process.env.TELEMETRY_UNIT_TEST_OUTPUT || "";
 
   const prompt = [
     "## Instructions",
     agentInstructions,
     "",
-    ...(targetIssueSection ? [targetIssueSection, ""] : []),
+    // W7: Multiple target issues
+    ...(targetIssueSections.length > 0 ? [
+      `## Target Issues (${targetIssueSections.length})`,
+      ...targetIssueSections.map((s) => s + "\n"),
+      targetIssueSections.length > 1
+        ? "**Resolve as many of these issues as you can in this session. Address them all if possible.**"
+        : "**Focus your transformation on resolving this specific issue.**",
+      "",
+    ] : []),
     "## Mission",
     mission,
     "",
@@ -143,12 +180,47 @@ export async function transform(context) {
       "Reference documents available in `library/` (use read_file for full content):",
       libraryIndex,
     ] : []),
+    // W9: worktree file listing
+    ...(worktreeFiles.length > 0 ? [
+      "",
+      `## Worktree Files (${worktreeFiles.length} non-ignored files)`,
+      worktreeFiles.join("\n"),
+    ] : []),
+    // W19: Current test state from telemetry
+    ...(telemetryTestSummary ? [
+      "",
+      "## Current Test State (from telemetry)",
+      `Summary: ${telemetryTestSummary}`,
+      ...(telemetryTestOutput ? [`\`\`\`\n${telemetryTestOutput}\n\`\`\``] : []),
+    ] : []),
+    // W17: Implementation review
+    ...(reviewAdvice ? [
+      "",
+      "## Implementation Review",
+      `**Completeness:** ${reviewAdvice}`,
+      ...((() => {
+        try {
+          const gaps = JSON.parse(reviewGapsRaw || "[]");
+          if (gaps.length > 0) {
+            return [
+              "",
+              "**Gaps Found:**",
+              ...gaps.map((g) => `- [${g.severity}] ${g.element}: ${g.description} (${g.gapType})`),
+              "",
+              "Address these gaps in your transformation if they fall within the target issues.",
+            ];
+          }
+        } catch { /* ignore */ }
+        return [];
+      })()),
+    ] : []),
     "",
     "## Your Task",
     "Analyze the mission and open issues (use list_issues tool).",
     "Read the source files you need (use read_file tool).",
-    "Determine the single most impactful next step to transform this repository.",
-    "Then implement that step, writing files and running run_tests to verify.",
+    issueNumbers.length > 1
+      ? "Resolve all target issues listed above. Implement all changes, write tests, update the website, and run run_tests to verify."
+      : "Determine the single most impactful next step to transform this repository.\nThen implement that step, writing files and running run_tests to verify.",
     "",
     "## When NOT to make changes",
     "If the existing code already satisfies all requirements in MISSION.md and all open issues have been addressed:",
@@ -162,6 +234,9 @@ export async function transform(context) {
     `- Run \`${testCommand}\` via run_tests to validate your changes`,
     "- Use list_issues to see open issues, get_issue for full details",
     "- Use read_file to read source files you need (don't guess at contents)",
+    ...(config.coverageGoals ? [
+      `- Required code coverage: ≥${config.coverageGoals.minLineCoverage}% lines, ≥${config.coverageGoals.minBranchCoverage}% branches`,
+    ] : []),
   ].join("\n");
 
   core.info(`Transform lean prompt length: ${prompt.length} chars`);
@@ -179,11 +254,61 @@ export async function transform(context) {
   const systemPrompt =
     "You are an autonomous code transformation agent. Your goal is to advance the repository toward its mission by making the most impactful change possible in a single step." + NARRATIVE_INSTRUCTION;
 
-  // ── Create custom tools (GitHub API + git) ─────────────────────────
+  // ── Create custom tools (GitHub API + git + W8 behaviour dry-run) ──
   const createTools = (defineTool, _wp, logger) => {
     const ghTools = createGitHubTools(octokit, repo, defineTool, logger);
     const gitTools = createGitTools(defineTool, logger);
-    return [...ghTools, ...gitTools];
+
+    // W8: Dry-run behaviour test tool — reads test specs and source code,
+    // returns them to the LLM for reasoning about whether code would pass
+    const dryRunBehaviourTests = defineTool("dry_run_behaviour_tests", {
+      description: "Read behaviour test specifications and the source code they test, then return both for analysis. Use this to check if your code changes would pass behaviour tests without running Playwright. Call this after making code changes but before committing.",
+      parameters: { type: "object", properties: {}, required: [] },
+      handler: async () => {
+        const behaviourPath = config.paths.behaviour?.path || "tests/behaviour/";
+        const sourcePath = config.paths.source?.path || "src/lib/";
+        const webPath = config.paths.web?.path || "src/web/";
+
+        const readDir = (dir, exts) => {
+          if (!existsSync(dir)) return [];
+          try {
+            return readdirSync(dir)
+              .filter((f) => exts.some((e) => f.endsWith(e)))
+              .slice(0, 10)
+              .map((f) => {
+                try {
+                  return { file: f, content: readFileSync(join(dir, f), "utf8") };
+                } catch { return { file: f, content: "(unreadable)" }; }
+              });
+          } catch { return []; }
+        };
+
+        const specs = readDir(behaviourPath, [".spec.js", ".spec.ts", ".test.js", ".test.ts"]);
+        const sources = readDir(sourcePath, [".js", ".ts"]);
+        const webFilesLocal = readDir(webPath, [".html", ".js"]);
+
+        if (specs.length === 0) {
+          return { textResultForLlm: "No behaviour test files found. Behaviour tests are not configured for this project." };
+        }
+
+        const parts = [
+          "## Behaviour Test Specifications",
+          ...specs.map((s) => `### ${s.file}\n\`\`\`\n${s.content}\n\`\`\``),
+          "",
+          "## Source Code Under Test",
+          ...sources.map((s) => `### ${s.file}\n\`\`\`\n${s.content}\n\`\`\``),
+        ];
+        if (webFilesLocal.length > 0) {
+          parts.push("", "## Website Files", ...webFilesLocal.map((s) => `### ${s.file}\n\`\`\`\n${s.content}\n\`\`\``));
+        }
+        parts.push("", "## Your Analysis", "Analyze whether the current source code and website would pass these behaviour tests. Report any gaps.");
+
+        logger.info(`[tool] dry_run_behaviour_tests: ${specs.length} specs, ${sources.length} sources, ${webFilesLocal.length} web files`);
+        return { textResultForLlm: parts.join("\n") };
+      },
+    });
+
+    return [...ghTools, ...gitTools, dryRunBehaviourTests];
   };
 
   // ── Run hybrid session ─────────────────────────────────────────────

package/src/copilot/config.js

@@ -92,6 +92,13 @@ function parseTuningProfile(profileSection) {
     issuesScan: profileSection["max-issues"] || 20,
     staleDays: profileSection["stale-days"] || 30,
     discussionComments: profileSection["max-discussion-comments"] || 10,
+    sessionTimeoutMs: profileSection["session-timeout-ms"] || 480000,
+    maxTokens: profileSection["max-tokens"] || 200000,
+    maxReadChars: profileSection["max-read-chars"] || 20000,
+    maxTestOutput: profileSection["max-test-output"] || 4000,
+    maxFileListing: profileSection["max-file-listing"] ?? 30,
+    maxLibraryIndex: profileSection["max-library-index"] || 2000,
+    maxFixTestOutput: profileSection["max-fix-test-output"] || 8000,
   };
 }
 
@@ -132,7 +139,7 @@ function readPackageJson(tomlPath, depsRelPath) {
  * @param {Object} [profilesSection] - The [profiles] section from TOML (source of truth)
  */
 function resolveTuning(tuningSection, profilesSection) {
-  const profileName = tuningSection.profile || "recommended";
+  const profileName = tuningSection.profile || "med";
   const tomlProfile = profilesSection?.[profileName];
   const profile = parseTuningProfile(tomlProfile) || FALLBACK_TUNING;
   const tuning = { ...profile, profileName };
@@ -149,6 +156,13 @@ function resolveTuning(tuningSection, profilesSection) {
     "max-issues": "issuesScan",
     "stale-days": "staleDays",
     "max-discussion-comments": "discussionComments",
+    "session-timeout-ms": "sessionTimeoutMs",
+    "max-tokens": "maxTokens",
+    "max-read-chars": "maxReadChars",
+    "max-test-output": "maxTestOutput",
+    "max-file-listing": "maxFileListing",
+    "max-library-index": "maxLibraryIndex",
+    "max-fix-test-output": "maxFixTestOutput",
   };
   for (const [tomlKey, jsKey] of Object.entries(numericOverrides)) {
     if (tuningSection[tomlKey] > 0) tuning[jsKey] = tuningSection[tomlKey];
@@ -239,6 +253,13 @@ export function loadConfig(configPath) {
   const execution = toml.execution || {};
   const bot = toml.bot || {};
 
+  // W13: Code coverage goals
+  const goals = toml.goals || {};
+  const coverageGoals = {
+    minLineCoverage: goals["min-line-coverage"] ?? 50,
+    minBranchCoverage: goals["min-branch-coverage"] ?? 30,
+  };
+
   // Mission-complete thresholds (with safe defaults)
   // C6: Removed minDedicatedTests and requireDedicatedTests
   const mc = toml["mission-complete"] || {};
@@ -267,6 +288,7 @@ export function loadConfig(configPath) {
     init: toml.init || null,
     tdd: toml.tdd === true,
     missionCompleteThresholds,
+    coverageGoals,
     maxTokensPerMaintain: resolvedLimits.maxTokensPerMaintain || 200000,
     writablePaths,
     readOnlyPaths,

package/src/copilot/copilot-session.js

@@ -74,7 +74,7 @@ export async function runCopilotSession({
   model = "gpt-5-mini",
   githubToken,
   tuning = {},
-  timeoutMs = 600000,
+  timeoutMs,
   agentPrompt,
   userPrompt,
   writablePaths,
@@ -94,6 +94,11 @@ export async function runCopilotSession({
 
   const wsPath = resolve(workspacePath);
 
+  // W11: Session timeout — defaults to 480s (8 min), leaving 2 min headroom
+  // below the 10-min workflow step timeout for graceful shutdown.
+  // Callers can override via timeoutMs parameter or tuning.sessionTimeoutMs.
+  const effectiveTimeoutMs = timeoutMs || tuning.sessionTimeoutMs || 480000;
+
   // ── Writable paths ──────────────────────────────────────────────────
   // Default: entire workspace is writable (local CLI mode)
   const effectiveWritablePaths = writablePaths || [wsPath + "/"];
@@ -154,7 +159,7 @@ export async function runCopilotSession({
   const systemPrompt = basePrompt + NARRATIVE_INSTRUCTION;
 
   // ── Session config ─────────────────────────────────────────────────
-  logger.info(`[agentic-lib] Creating session (model=${model}, workspace=${wsPath})`);
+  logger.info(`[agentic-lib] Creating session (model=${model}, workspace=${wsPath}, timeout=${Math.round(effectiveTimeoutMs / 1000)}s)`);
 
   const client = new CopilotClient({
     env: { ...process.env, GITHUB_TOKEN: copilotToken, GH_TOKEN: copilotToken },
@@ -192,7 +197,7 @@ export async function runCopilotSession({
         // Truncate large read_file results to prevent context overflow
         if (input.toolName === "read_file" || input.toolName === "view") {
           const resultText = input.toolResult?.textResultForLlm || "";
-          const MAX_READ_CHARS = 20000;
+          const MAX_READ_CHARS = tuning.maxReadChars || 20000;
           if (resultText.length > MAX_READ_CHARS) {
             hookOutput.modifiedResult = {
               ...input.toolResult,
@@ -296,7 +301,7 @@ export async function runCopilotSession({
 
   const prompt = userPrompt || [
     `# Mission\n\n${missionText}`,
-    `# Current test state\n\n\`\`\`\n${initialTestOutput.substring(0, 4000)}\n\`\`\``,
+    `# Current test state\n\n\`\`\`\n${initialTestOutput.substring(0, tuning.maxTestOutput || 4000)}\n\`\`\``,
     "",
     "Implement this mission. Read the existing source code and tests,",
     "make the required changes, run run_tests to verify, and iterate until all tests pass.",
@@ -315,7 +320,7 @@ export async function runCopilotSession({
   }
   for (let attempt = 0; attempt <= maxRetries; attempt++) {
     try {
-      response = await session.sendAndWait(sendOptions, timeoutMs);
+      response = await session.sendAndWait(sendOptions, effectiveTimeoutMs);
       break;
     } catch (err) {
       if (isRateLimitError(err) && attempt < maxRetries) {

package/src/iterate.js

@@ -108,7 +108,7 @@ export function readTransformationCost(targetPath) {
 
 /**
  * Read transformation budget from agentic-lib.toml.
- * Falls back to 8 (the "recommended" profile default).
+ * Falls back to 8 (the "med" profile default).
  */
 export function readBudget(targetPath) {
   const tomlPath = resolve(targetPath, "agentic-lib.toml");

package/src/seeds/zero-package.json

@@ -17,7 +17,7 @@
   "author": "",
   "license": "MIT",
   "dependencies": {
-    "@xn-intenton-z2a/agentic-lib": "^7.4.23"
+    "@xn-intenton-z2a/agentic-lib": "^7.4.25"
   },
   "devDependencies": {
     "@playwright/test": "^1.58.0",