You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Sync the pr-review-mention.yml thin-caller stub to the org standard template by replacing the explicit secrets: block with secrets: inherit, eliminating the drift that causes ~12% of workflow runs to fail.
Root Cause
.github/workflows/pr-review-mention.yml passes secrets explicitly to the reusable:
The org standard template (petry-projects/.github/standards/workflows/pr-review-mention.yml) uses secrets: inherit, which was the intended form when the stub was built. Explicit secret expansion causes action_required outcomes when the calling context cannot expand one or both secrets (e.g., triggered by pull_request_review_comment or pull_request events from certain actors). action_required conclusions are counted as failures by the Fleet Monitor.
Implementation Checklist
tests/test_pr_review_mention_stub.py β new compliance test that asserts secrets: inherit is used (TDD: red first)
.github/workflows/pr-review-mention.yml β replace explicit secrets: block with secrets: inherit; also sync the Requires: comment to match the org standard
Tests to Write
tests/test_pr_review_mention_stub.py β reads the local YAML, checks jobs.pr-review-mention.secrets == 'inherit', fails if explicit mapping is present
The org standard template (petry-projects/.github/standards/workflows/pr-review-mention.yml) uses secrets: inherit. Explicit secret expansion triggers action_required outcomes when the calling context cannot expand one or both secrets (e.g., for pull_request_review_comment or review_requested events from certain actors). action_required runs are counted as failures by the Fleet Monitor.
Test Results
# Compliance test (new)
python3 tests/test_pr_review_mention_stub.py
OK: uses = 'petry-projects/.github/.github/workflows/pr-review-mention-reusable.yml@v2'
OK: secrets: inherit
PASS: .github/workflows/pr-review-mention.yml conforms to the org standard stub contract.
# Unit tests
bash tests/test_copilot_chat.sh β Results: 13 passed, 0 failed
bash tests/test_list_prs_sort.sh β Results: 17 passed, 0 failed
# Bats suite (fleet_report, token_report, model_pricing, test_review_cycle)
1..93 (93 tests, 0 failures)
# dev-lead-lint.sh
shellcheck on 34 file(s)... PASS
validating 4 agent profile(s)... PASS
[lint] all checks passed
Files Changed
.github/workflows/pr-review-mention.yml: replaced explicit secrets: block with secrets: inherit; synced Requires: comment to match org standard
tests/test_pr_review_mention_stub.py: new compliance test β validates the stub uses secrets: inherit
Notes
Triggers and job-level permissions were not changed (per the stub's AGENTS contract). The reusable itself (pr-review-mention-reusable.yml@v2) already handles empty/inherited secrets gracefully via its job-level if: condition.
β Auto-resolved by Fleet Monitor. This workflow's failure rate has been below the 10% threshold (or the workflow was removed) since 2026-06-04 (41 days) β the monitor stopped flagging it. Closing. It will reopen automatically if the workflow crosses the threshold again.
Activity
don-petry commentedon Jun 8, 2026
Dev-Lead Implementation Plan
Issue: #431 β [Fleet Monitor] petry-projects/.github-private β pr-review-mention.yml
Scope
Sync the
pr-review-mention.ymlthin-caller stub to the org standard template by replacing the explicitsecrets:block withsecrets: inherit, eliminating the drift that causes ~12% of workflow runs to fail.Root Cause
.github/workflows/pr-review-mention.ymlpasses secrets explicitly to the reusable:The org standard template (
petry-projects/.github/standards/workflows/pr-review-mention.yml) usessecrets: inherit, which was the intended form when the stub was built. Explicit secret expansion causesaction_requiredoutcomes when the calling context cannot expand one or both secrets (e.g., triggered bypull_request_review_commentorpull_requestevents from certain actors).action_requiredconclusions are counted as failures by the Fleet Monitor.Implementation Checklist
tests/test_pr_review_mention_stub.pyβ new compliance test that assertssecrets: inheritis used (TDD: red first).github/workflows/pr-review-mention.ymlβ replace explicitsecrets:block withsecrets: inherit; also sync theRequires:comment to match the org standardTests to Write
tests/test_pr_review_mention_stub.pyβ reads the local YAML, checksjobs.pr-review-mention.secrets == 'inherit', fails if explicit mapping is presentTest Command
python3 tests/test_pr_review_mention_stub.pydon-petry commentedon Jun 8, 2026
Dev-Lead: Implementation Complete
Plan Execution
secrets:block diverged from org standard (secrets: inherit)tests/test_pr_review_mention_stub.py(TDD red).github/workflows/pr-review-mention.ymlto usesecrets: inherit(TDD green)scripts/dev-lead-lint.sh(all pass)Root Cause
The local stub used an explicit
secrets:block:The org standard template (
petry-projects/.github/standards/workflows/pr-review-mention.yml) usessecrets: inherit. Explicit secret expansion triggersaction_requiredoutcomes when the calling context cannot expand one or both secrets (e.g., forpull_request_review_commentorreview_requestedevents from certain actors).action_requiredruns are counted as failures by the Fleet Monitor.Test Results
Files Changed
.github/workflows/pr-review-mention.yml: replaced explicitsecrets:block withsecrets: inherit; syncedRequires:comment to match org standardtests/test_pr_review_mention_stub.py: new compliance test β validates the stub usessecrets: inheritNotes
Triggers and job-level permissions were not changed (per the stub's AGENTS contract). The reusable itself (
pr-review-mention-reusable.yml@v2) already handles empty/inherited secrets gracefully via its job-levelif:condition.feat: implement issue #431 β [Fleet Monitor] petry-projects/.github-pβ¦
feat: implement issue #431 β [Fleet Monitor] petry-projects/.github-pβ¦
feat: implement issue #431 β [Fleet Monitor] petry-projects/.github-pβ¦
feat: implement issue #431 β [Fleet Monitor] petry-projects/.github-pβ¦
don-petry commentedon Jul 15, 2026
β Auto-resolved by Fleet Monitor. This workflow's failure rate has been below the 10% threshold (or the workflow was removed) since 2026-06-04 (41 days) β the monitor stopped flagging it. Closing. It will reopen automatically if the workflow crosses the threshold again.