Repository navigation
Description
Repin this repo's dev-lead caller stub from the orphaned dev-lead/v1-* channel family to dev-lead/v139-*.
The change
One file: .github/workflows/dev-lead.yml. Two lines.
- uses: petry-projects/.github-private/.github/workflows/dev-lead-reusable.yml@dev-lead/v1-ring0
+ uses: petry-projects/.github-private/.github/workflows/dev-lead-reusable.yml@dev-lead/v139-ring0
with:
- agent_ref: dev-lead/v1-ring0
+ agent_ref: dev-lead/v139-ring0Match by content, not line number. Both must change and must stay equal — a mismatch means the reusable and the scripts it checks out come from different channels.
Change nothing else. on:, permissions:, secrets: and concurrency: are not repo-adjustable in this stub.
⚠️ This one is a real version move, not a rename
dev-lead/v1-ring0 -> 6d3623187cfe (2026-07-10)
dev-lead/v139-ring0 -> 40f6c2325ef7 (2026-08-18)
This repo currently executes dev-lead code from 2026-07-10 — 152 commits behind. After this change it runs 40f6c232 (release v139.2.0), the commit petry-projects/.github-private has run in production since 2026-08-18. It moves forward only; nothing regresses.
Among the improvements this pulls in: scripts/engine.sh gained a guard that skips the Copilot fallback when COPILOT_GITHUB_TOKEN is a classic ghp_ PAT, treating it as a configuration gap rather than a hard failure. Without it, every Claude session-limit blip in this repo dies at Error: Classic Personal Access Tokens (ghp_) are not supported by Copilot instead of falling through to Gemini.
Why
dev-lead is on major 139 (highest release v139.5.0). _resolved_channel derives v<major>-<tier>, so promotion advances the v139 line — while every caller stub still pins a v1-* tag that nothing moves. The v1-* family was orphaned when ~137 spurious major bumps (v2.0.0 07-20 → v139.0.0 08-12) walked the major forward. Root cause: petry-projects/.github-private#1592.
Tiers were deliberately flattened to 40f6c232 so no repo regresses on cutover. petry-projects/.github-private is already repinned this way (PR #1653, merged).
Acceptance criteria
-
.github/workflows/dev-lead.ymlpinsdev-lead/v139-ring0in bothuses:andagent_ref:. - No other line in the file changes.
- No other file changes.
- Add the missing SonarCloud marker. This repo's stub is the odd one out — it lacks the
# NOSONAR(githubactions:S7637) first-party channel refinline comment thatpetry-projects/.github-private's stub carries on itsuses:line.standards/ci-standards.mdstates that marker is required on a channel-pinned first-partyuses:or SonarCloud'sgithubactions:S7637fails the Quality Gate, and this repo is SonarCloud-gated. Add it as part of this change and confirm the SonarCloud check passes on the PR. - The workflow still parses as valid YAML.
Do not
- Do not repin to
dev-lead/v139-next. The tier is preserved exactly:ring0→ring0. - Do not change a test that asserts a channel tier invariant. If a test hardcodes the literal
v1-ring0, update that literal only.
Refs petry-projects/.github-private#1592, petry-projects/.github#657, petry-projects/.github-private#1495
NOTE — this stub has no comment block to update
Unlike the other five repos, this stub is a shorter variant: the uses: line sits at ~L46 (not ~L64) and it carries no explanatory comment block naming the channel. So there is no prose to correct here — only the two pin lines, plus the missing NOSONAR(githubactions:S7637) marker already listed above.
That structural difference is itself drift from the canonical stub (which standards/ci-standards.md describes as full-file identical across adopting repos, modulo the ring/channel pin). Do not try to reconcile the whole file in this PR — that is a separate change. Restrict this PR to the two pin lines and the NOSONAR marker.
UPDATE 2026-09-07 — the numbers moved, and this repo is now the fleet's bootstrap canary
The target commit has advanced since this issue was written. Promotion ran while this sat unreleased, so v139-ring0 no longer points at 40f6c232:
dev-lead/v1-ring0 -> 6d3623187cfe (2026-07-10) <- what this repo runs today
dev-lead/v139-ring0 -> 90dcc01ca854 (2026-09-03) <- what it will run after this change
+194 commits, not the +152 first stated
90dcc01c is also main, v139-next and v139-ring1 — the whole upper pipeline has converged there.
The change itself is unchanged: uses: and agent_ref: both dev-lead/v1-ring0 → dev-lead/v139-ring0, plus the missing NOSONAR(githubactions:S7637) marker. Only the resulting version is bigger than originally described.
Why this repo, and why now
The dev-lead canary tiers are completely unoccupied:
v139-next occupants: NONE
v139-ring0 occupants: NONE
v139-ring1 occupants: NONE
v139-stable occupants: .github-private, markets, google-app-scripts, ContentTwin
Every repo sits on stable or a stale v1-* tag, so no candidate ever accumulates real soak evidence. The ring1->stable gate currently reports dwell=88h/12h, sample=196/1 — impressive numbers measured entirely on repos running other versions. Nothing in the fleet executes the candidate.
Working out who can occupy each tier:
next— members[$host]=petry-projects/.github-privateonly, and SC2 (petry-projects/.github-private#1624) requires that repo's stub to pin a stable tier so a broken candidate cannot block its own fix.nextis therefore structurally unoccupiable for dev-lead.ring0— members[$org_infra]=petry-projects/.github+petry-projects/.github-private. The latter is pinned stable by SC2.
So this repo is the only viable first canary in the fleet. This change is what bootstraps the pipeline; until it lands, every promotion decision for dev-lead rests on evidence about versions nobody runs.
Risk, stated plainly
This is a +194-commit jump onto code with no prior canary coverage. That is the actual cost, and it is accepted deliberately: taking the first risk is what ring0 is for, this repo generates evidence quickly (44 dev-lead runs since the current cut), and the SC2 invariant means petry-projects/.github-private stays on stable with a working dev-lead to ship a fix if this canary breaks.
TalkTerm and bmad-bgreat-suite (ring1) follow after this repo has been observed running clean — not in parallel.
Refs petry-projects/.github-private#1592, petry-projects/.github-private#1624, #1049, #1066
Activity
don-petry commentedon Sep 7, 2026
Dev-Lead Implementation Plan
Issue: #1063 — Repin dev-lead caller stub: dev-lead/v1-ring0 -> dev-lead/v139-ring0 (orphaned channel family, 152 commits behind, #1592)
Scope
Repin this repo's
dev-leadcaller stub (.github/workflows/dev-lead.yml) from the orphaneddev-lead/v1-ring0channel todev-lead/v139-ring0, and add the missing# NOSONAR(githubactions:S7637)inline marker required bystandards/ci-standards.mdon the channel-pinned first-partyuses:line.Implementation Checklist
.github/workflows/dev-lead.yml—uses:ref:@dev-lead/v1-ring0→@dev-lead/v139-ring0, and append# NOSONAR(githubactions:S7637) first-party channel ref.github/workflows/dev-lead.yml—agent_ref:input:dev-lead/v1-ring0→dev-lead/v139-ring0(must stay equal to theuses:channel)on:/permissions:/concurrency:/secrets:untouchedTests to Write
dev-lead-stub-channel-sync.batsfixtures usev1-ring0as generic parsing inputs asserting the agent_ref↔uses sync invariant — they do not assert this repo's tier, so per the issue's "do not change a test that asserts a channel tier invariant" they are left untouched.Verification
python3 -c yaml.safe_load/ actionlint if available)bash .dev-lead/scripts/dev-lead-lint.shdev-lead/v139-ring0githubactions:S7637check passes on the PR (marker now present)Test Command
bash .dev-lead/scripts/dev-lead-lint.sh(repo lint); no unit-test suite applies to this config changedon-petry commentedon Sep 7, 2026
Dev-Lead: Implementation Complete
Plan Execution
uses:ref@dev-lead/v1-ring0→@dev-lead/v139-ring0agent_ref:inputdev-lead/v1-ring0→dev-lead/v139-ring0(kept equal touses:channel — no split-brain)# NOSONAR(githubactions:S7637) first-party channel refinline marker on the channel-pinned first-partyuses:line (required bystandards/ci-standards.mdon SonarCloud-gated repos)on:,permissions:,concurrency:,secrets:left untouchedring0→ring0(not repinned tov139-next)Diff
Verification
yaml.safe_load— jobdev-leadpresent)useschannel =v139-ring0,agent_refchannel =v139-ring0— equal & correctbash .dev-lead/scripts/dev-lead-lint.sh): all checks passed (shellcheck on 35 files + CODEOWNERS validation)git status --porcelainshows only.github/workflows/dev-lead.ymlmodifiedFiles Changed
.github/workflows/dev-lead.yml: two pin lines repinnedv1-ring0→v139-ring0; added theNOSONAR(githubactions:S7637)inline marker to theuses:line.Notes
test/scripts/compliance-audit/dev-lead-stub-channel-sync.batsfixtures usev1-ring0as generic parsing inputs asserting the agent_ref↔uses sync invariant; they do not assert this repo's actual tier, so per the issue's "do not change a test that asserts a channel tier invariant" they were left untouched.githubactions:S7637Quality Gate check should now pass on the PR — the required inline marker is present on the first-party channel-pinneduses:line.dev-lead/v139-ring0(90dcc01cper the 2026-09-07 update); nothing regresses.feat: implement issue #1063 — Repin dev-lead caller stub: dev-lead/v1…