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
The template-drift check is RED on new PRs here (e.g. #1324), because three
seeded template stubs diverge from their standards-derived baseline in
petry-projects/.github:
Stub
Cause of drift
.github/workflows/pr-review-mention.yml
#1317 renamed the secret in the stub (GH_PAT_WORKFLOWS→GH_PAT_DON_PETRY) but not in the standards source. Fixed upstream by combsmith/.github#791 — MERGE THAT FIRST.
.github/workflows/copilot-setup-steps.yml
standards updated upstream; seeded stub is stale.
.github/workflows/sonarcloud.yml
standards updated upstream; seeded stub is stale.
Per the canary channel rule, the pr-review agent skips approval on any red CI,
so this one drift stalls dev-lead auto-merge fleet-wide. (Verified: pre-#1317
PR #1292 passes template-drift; post-#1317 PR #1324 fails on exactly these 3.)
Change
Re-seed the drifted stubs from the canonical standards — do not hand-edit the
stubs (the check explicitly forbids it; edit standards/ in .github instead):
scripts/seed-repo-template.sh # STANDARDS_REPO defaults to petry-projects/.github
Commit the regenerated stubs. Only the three files above should change.
Ordering (hard dependency)
Must land only AFTER combsmith/.github#791 merges. If re-seeded first, the
regenerated pr-review-mention.yml will revert to GH_PAT_WORKFLOWS, undoing the #1317 rename.
Sanity-check the copilot-setup-steps / sonarcloud changes are the intended
upstream standards updates, not a revert of an intentional local customization.
Fails closed: a re-seed that cannot fetch a standard must exit non-zero, never
emit a partial/green result.
Thanks for the thorough write-up and for tracking down the exact ordering constraint — this is a clear and well-scoped report. Before we proceed with the re-seed, could you confirm whether combsmith/.github#791 has already merged? That's the hard prerequisite, and knowing its current status will determine whether we can act on this immediately or need to wait.
But re-seeding copilot-setup-steps.yml and sonarcloud.yml would drag actions/checkout pins backward (e.g. sonarcloud stub v6.0.3 → standards v6.0.2; a v7→v6 revert on repo-template), reverting Dependabot bumps — which fails this issue's own "don't revert an intentional customization" acceptance criterion and won't durably fix the drift (Dependabot re-bumps).
The durable fix is the reverse direction: bring the .githubstandards (and repo-template's Dependabot config) up to the current pinned checkout versions, so a subsequent re-seed aligns without downgrade. That's an org decision on checkout-version adoption + a cross-repo change, not a blanket re-seed. Removing the dev-lead label to stop no-op re-triggering until the direction is set.
Note: the identity-fix half this issue was spun off to support is validated independently (dev-lead PR #1324 commits authored by don-petry). This residual template-drift is a pre-existing standards↔stub version skew, not a #1316/#1317 regression.
So re-seeding now will not revert the #1317 rename — the ordering guard is clear.
2. Drift confirmed (exactly the 3 named stubs)
bash scripts/template_stub_drift.sh → copilot-setup-steps.yml, pr-review-mention.yml, sonarcloud.yml
are DRIFTED vs petry-projects/repo-template; every other covered file is ALIGNED.
3. Sanity-check (acceptance criterion) — a downgrade is hiding in the diff ⚠️
The seed emission is regenerated from standards/, and standards are behind the template on actions/checkout:
Stub
Re-seed effect
pr-review-mention.yml
Adds GH_PAT_DON_PETRY || GH_PAT_WORKFLOWS — the intended fix ✅
sonarcloud.yml
Adds a genuine "Set up JDK 21" step ✅ but alsoactions/checkoutv7.0.0 → v6.0.3 ⬇️
copilot-setup-steps.yml
Only change is actions/checkoutv7.0.0 → v6.0.2 ⬇️ — no forward standards update at all
repo-template runs its own Dependabot (open PR #62), so the committed v7.0.0 pins are almost certainly
Dependabot bumps in the template. Re-seeding as-written would silently revert those bumps / downgrade a
pinned action fleet-wide — the "revert of an intentional local change" the acceptance criteria call out.
4. Structural note — this is a cross-repo write, not an in-repo edit
template_stub_drift.sh reads committed blobs from petry-projects/repo-template via the contents API.
No file in .github-private clears this drift; the fix is a re-seed PR against repo-template
(seed-repo-template.sh → branch + contents PUT + gh pr create), which is outside the dev-lead's
"don't open PRs / CI handles git" flow (CI here only covers this repo).
5. Fails-closed — already holds ✅
seed-repo-template.sh exits non-zero if any standard can't be fetched
(_emit_workflow/_emit_baseline return 1, propagated by _seed_repo) — no partial/green emission is possible.
Recommendation
Do not blindly re-seed — it downgrades actions/checkout in 2 of the 3 stubs. Preferred sequence (no downgrade):
Bump actions/checkout to the v7.0.0 SHA (9c091bb…) in petry-projects/.github standards/workflows/{copilot-setup-steps,sonarcloud}.yml, and merge.
→ copilot-setup-steps.yml drift then clears with no template change; sonarcloud.yml keeps v7.0.0 and
only gains JDK-21.
Then re-seed repo-template (bash scripts/seed-repo-template.sh) so pr-review-mention.yml gains the
secret fallback and sonarcloud.yml gains JDK-21 — all forward, zero downgrade.
Alternative (as literally worded in this issue): re-seed now and accept the v7→v6 checkout downgrade as
"standards is canonical". This unblocks the fleet immediately but reverts the template's Dependabot bumps.
Requesting a human go/no-go on which path, plus authorization for the cross-repo write (a PR on petry-projects/repo-template, and for the preferred path a PR on petry-projects/.github) — neither of which
the dev-lead performs autonomously under its constraints.
Superseded — closing. dev-lead's investigation showed the template-drift check guards petry-projects/repo-template's stubs (not this repo's), and repo-template is AHEAD on actions/checkout (v7.0.0, Dependabot). The re-seed can't happen here. Split into the correct no-downgrade fix:
Activity
github-actions commentedon Jul 19, 2026
Thanks for the thorough write-up and for tracking down the exact ordering constraint — this is a clear and well-scoped report. Before we proceed with the re-seed, could you confirm whether combsmith/.github#791 has already merged? That's the hard prerequisite, and knowing its current status will determine whether we can act on this immediately or need to wait.
don-petry commentedon Jul 19, 2026
Blocked on a maintainer decision — dev-lead correctly declined to auto-implement.
dev-lead ran (commit identity resolved to
don-petry✅ — the #1316 fix is live here) and made no changes, with sound reasoning:GH_PAT_DON_PETRY || GH_PAT_WORKFLOWS.copilot-setup-steps.ymlandsonarcloud.ymlwould dragactions/checkoutpins backward (e.g. sonarcloud stubv6.0.3→ standardsv6.0.2; av7→v6revert on repo-template), reverting Dependabot bumps — which fails this issue's own "don't revert an intentional customization" acceptance criterion and won't durably fix the drift (Dependabot re-bumps).The durable fix is the reverse direction: bring the
.githubstandards (and repo-template's Dependabot config) up to the current pinned checkout versions, so a subsequent re-seed aligns without downgrade. That's an org decision on checkout-version adoption + a cross-repo change, not a blanket re-seed. Removing thedev-leadlabel to stop no-op re-triggering until the direction is set.Note: the identity-fix half this issue was spun off to support is validated independently (dev-lead PR #1324 commits authored by don-petry). This residual template-drift is a pre-existing standards↔stub version skew, not a #1316/#1317 regression.
don-petry commentedon Jul 19, 2026
Re-scoped prerequisite: combsmith/.github#793.
Corrected plan after diffing the stubs against standards:
sonarcloud.yml: stub is AHEAD (adds a JDK-21 setup-java step + checkout v6.0.3). Blind re-seed would revert both → 💡 Opus 4.8 Engine Upgrade: Dynamic Workflows + 4x Honesty for Review Quality #793 adopts them into standards first.copilot-setup-steps.yml: only drift is doc-comments the stub stripped (checkout identical v6.0.2) → re-seeds safely, no standards change needed.pr-review-mention.yml: re-seeds cleanly (standards fixed by 💡 Stacked PR Awareness for Review and Dev-Lead Pipelines #792).So once #793 merges, this re-seed becomes safe on all three. I'll re-add the
dev-leadlabel then.don-petry commentedon Jul 19, 2026
Dev-Lead: Investigation complete — re-seed is a cross-repo action with a downgrade caveat (needs go/no-go)
1. Hard dependency IS satisfied ✅
#791is an issue; its implementing PRpetry-projects/.github#792merged 2026-07-19 14:21 UTC. Thecanonical source now emits the retained fallback:
So re-seeding now will not revert the #1317 rename — the ordering guard is clear.
2. Drift confirmed (exactly the 3 named stubs)
bash scripts/template_stub_drift.sh→copilot-setup-steps.yml,pr-review-mention.yml,sonarcloud.ymlare DRIFTED vs
petry-projects/repo-template; every other covered file is ALIGNED.3. Sanity-check (acceptance criterion) — a downgrade is hiding in the diff⚠️
The seed emission is regenerated from
standards/, and standards are behind the template onactions/checkout:pr-review-mention.ymlGH_PAT_DON_PETRY || GH_PAT_WORKFLOWS— the intended fix ✅sonarcloud.ymlactions/checkoutv7.0.0 → v6.0.3 ⬇️copilot-setup-steps.ymlactions/checkoutv7.0.0 → v6.0.2 ⬇️ — no forward standards update at allrepo-templateruns its own Dependabot (open PR #62), so the committed v7.0.0 pins are almost certainlyDependabot bumps in the template. Re-seeding as-written would silently revert those bumps / downgrade a
pinned action fleet-wide — the "revert of an intentional local change" the acceptance criteria call out.
4. Structural note — this is a cross-repo write, not an in-repo edit
template_stub_drift.shreads committed blobs frompetry-projects/repo-templatevia the contents API.No file in
.github-privateclears this drift; the fix is a re-seed PR againstrepo-template(
seed-repo-template.sh→ branch + contents PUT +gh pr create), which is outside the dev-lead's"don't open PRs / CI handles git" flow (CI here only covers this repo).
5. Fails-closed — already holds ✅
seed-repo-template.shexits non-zero if any standard can't be fetched(
_emit_workflow/_emit_baselinereturn 1, propagated by_seed_repo) — no partial/green emission is possible.Recommendation
Do not blindly re-seed — it downgrades
actions/checkoutin 2 of the 3 stubs. Preferred sequence (no downgrade):actions/checkoutto the v7.0.0 SHA (9c091bb…) inpetry-projects/.githubstandards/workflows/{copilot-setup-steps,sonarcloud}.yml, and merge.→
copilot-setup-steps.ymldrift then clears with no template change;sonarcloud.ymlkeeps v7.0.0 andonly gains JDK-21.
repo-template(bash scripts/seed-repo-template.sh) sopr-review-mention.ymlgains thesecret fallback and
sonarcloud.ymlgains JDK-21 — all forward, zero downgrade.Alternative (as literally worded in this issue): re-seed now and accept the v7→v6
checkoutdowngrade as"standards is canonical". This unblocks the fleet immediately but reverts the template's Dependabot bumps.
Requesting a human go/no-go on which path, plus authorization for the cross-repo write (a PR on
petry-projects/repo-template, and for the preferred path a PR onpetry-projects/.github) — neither of whichthe dev-lead performs autonomously under its constraints.
don-petry commentedon Jul 19, 2026
Superseded — closing. dev-lead's investigation showed the
template-driftcheck guards petry-projects/repo-template's stubs (not this repo's), and repo-template is AHEAD onactions/checkout(v7.0.0, Dependabot). The re-seed can't happen here. Split into the correct no-downgrade fix:actions/checkout→ v7.0.0 (copilot-setup-steps + sonarcloud; corrects 💡 OWASP Agentic Top 10 Posture Scoring: Operationalize the Published Framework #794). Clears copilot-setup-steps drift on its own.Together these turn
template-driftGREEN with zero downgrades. Thanks to dev-lead for catching the repo-template target + downgrade twice.