Merge pull request #410 from githubnext/codex/evergreen-diagnostics-timeout

Improve Evergreen CI diagnostics

Author: mrjf

.github/workflows/evergreen.lock.yml

@@ -22,6 +22,8 @@ concurrency:
   group: gh-aw-${{ github.workflow }}-${{ github.event.inputs.pr || github.event.pull_request.number || github.ref || github.run_id }}
   cancel-in-progress: true
 
+timeout-minutes: 60
+
 permissions:
   contents: read
   issues: read
@@ -52,7 +54,7 @@ jobs:
         name: Evaluate PR gate state
         shell: bash
         env:
-          GH_TOKEN: ${{ github.token }}
+          GH_TOKEN: ${{ secrets.GH_AW_CI_TRIGGER_TOKEN || github.token }}
           REPO: ${{ github.repository }}
           EVENT_NAME: ${{ github.event_name }}
           EVENT_ACTION: ${{ github.event.action }}
@@ -434,11 +436,38 @@ pre-agent-steps:
       echo "Evergreen workspace is on branch $current_branch at $current_head_sha for PR #$PR_NUMBER."
 
 tools:
+  timeout: 600
   github:
     toolsets: [repos, issues, pull_requests, actions]
   bash:
-    - gh
+    - awk
+    - base64
+    - bun:*
+    - find
+    - gh:*
+    - git add:*
+    - git branch:*
+    - git checkout:*
+    - git commit:*
+    - git diff:*
+    - git log:*
+    - git rev-parse:*
+    - git rm:*
+    - git show:*
+    - git status
+    - git switch:*
+    - grep
     - jq
+    - mkdir
+    - node:*
+    - npm:*
+    - npx:*
+    - pwd
+    - rg
+    - rm:*
+    - sed
+    - tar:*
+    - unzip:*
 
 imports:
   - shared/skills/pr-intake.md
@@ -529,6 +558,10 @@ are merge gates.
 4. Run `ci-run-deduper`.
 5. Run `ci-gate-evaluator`.
 6. Run `ci-log-parser` for failing checks.
+   For CI, lint, typecheck, or test failures, collect the exact failing command
+   and full relevant diagnostics before guessing or delegating. Use GitHub job
+   logs/API, downloaded logs when needed, and targeted local reproduction such as
+   `bun run lint` or `bun run typecheck`.
 7. Run `merge-blocker-comment-reader` only for configured gates or explicit
    merge blockers.
 8. Run `deterministic-repair` before agentic edits.

.github/workflows/evergreen.md

@@ -98,8 +98,8 @@ readiness controller and the agentic orchestrator must both respect this file.
   through safe outputs by the agent (never from preflight) and labeled
   `evergreen: trigger CI`; they do not count as semantic repair attempts.
 - Token policy: `GITHUB_TOKEN` for reads and control-plane label writes.
-  `GH_AW_CI_TRIGGER_TOKEN` (existing PAT) is used by gh-aw for pushes so that PR
-  branch pushes trigger CI (default-token pushes do not).
+  `GH_AW_CI_TRIGGER_TOKEN` (existing PAT) is used by gh-aw for CI reruns and
+  pushes so that PR branch pushes trigger CI (default-token pushes do not).
 
 ## Repair Policy
 
@@ -121,6 +121,10 @@ readiness controller and the agentic orchestrator must both respect this file.
   - E2E: `bun run test:e2e`
   - Golden snapshots: `python golden/generate.py`
   - Workflow compile: `gh aw compile` (and `apm compile` when APM sources change)
+- CI/lint diagnosis policy: when a CI gate fails, fetch the exact failing job
+  logs and run the targeted repo command locally before editing. For lint
+  failures, `bun run lint` is the source of truth; do not guess from truncated
+  GitHub summaries.
 - Generated file policy: recompile committed lockfiles/snapshots when their
   sources change. After editing any `.github/workflows/*.md` workflow, recompile
   and commit the generated `.lock.yml`.