Repository navigation
Description
Story
As the org,
I want the weekly standards sweep to carry the persona-mention caller stub to the fleet,
so that a persona can be summoned somewhere other than petry-projects/.github-private.
Phase 2 of combsmith/.github-private#1643. This is the same shape as #1128 (#1045 Step D for apply-repo-settings): the template already exists, the stub is simply not in the deployable list.
Measured state, 2026-09-21
.github/workflows/persona-mention.yml exists in .github-private only — not in .github, TalkTerm, markets, ContentTwin, bmad-bgreat-suite, repo-template, broodly or incubator. DEPLOYABLE_WORKFLOWS in scripts/deploy-standard-workflows.sh today is: pr-review-mention, dev-lead, agent-shield, auto-rebase, dependabot-automerge, dependabot-rebase, dependency-audit, add-to-project, initiative-driver, pr-auto-review, feature-ideation. persona-mention.yml is absent, so no sweep can ever carry it.
standards/workflows/persona-mention.yml already exists, so nothing needs authoring — only listing.
BLOCKED BY #783 — do not start until persona-mention/v1-stable exists
Matching refs today: v1-next (fde249e0), v1-ring0 (a781ad9d), v1-ring1 (366ab877). There is no v1-stable.
A stable-tier repo (the * ring, i.e. most of the fleet) computes @persona-mention/v1-stable, and #1088's assert-exists check refuses a pin whose tag does not exist. Listing this stub before the ring soak completes therefore breaks the sweep for every stable-tier repo. #1128 records this precondition verbatim for the sibling agent; the same rule applies here.
Acceptance Criteria
- Re-verify the precondition in code, not just here:
persona-mention/v1-stableexists and the ring soak ([promote persona-mention 1/3] Soak through the rings (next → ring0 → ring1 → stable) #783) reached it without skipping a tier. - Add
persona-mention.ymltoDEPLOYABLE_WORKFLOWSinscripts/deploy-standard-workflows.sh. - A
--dry-runof the sweep for--workflow persona-mention.ymlshows every target repo getting a pin whose tag exists — stable-tier repos@persona-mention/v1-stable, ring members theirv1-<tier>. Paste that output into the PR. If any computed pin does not exist, stop and report it. - Do not touch the template's
uses:/checkout_ref:lockstep lines. - Tests: the list contains the entry, and ring-pin emission for
persona-mentionin a stable-tier repo resolves tov1-stable. shellcheck --severity=warning -xclean; tests registered in the existing bats job.
Dev Notes
- Deployment is PR-based and labelled
standards-sync; this change only makes the stub eligible. - Adoption is not the same as reach. A repo with the stub still routes only personas whose manifests enable the mention surface, and the opt-out labels are already provisioned fleet-wide ([qa-lead S1] Provision <id>:hands-off labels across the fleet — the opt-out is inert where the persona actually runs .github-private#1644).
- The sweep opens one PR per fleet repo with a full CI fan-out. Check the org automation-PR cap (
standards/pr-limits.json, currently 50 org-wide) before releasing, as #1045 Step D: add apply-repo-settings.yml to DEPLOYABLE_WORKFLOWS (Step C done — v1-stable exists) #1128 notes.
References
- Epic: qa-lead — close the gap between the declared persona and the served one .github-private#1643 (epic, Phase 2), [promote persona-mention 1/3] Soak through the rings (next → ring0 → ring1 → stable) #783 (the ring soak that gates this), #1045 Step D: add apply-repo-settings.yml to DEPLOYABLE_WORKFLOWS (Step C done — v1-stable exists) #1128 (the sibling Step D), RING_REUSABLES has drifted from canary-rings.json — 7 agents deploy verbatim, 2 with pins that do not exist #1088 (assert-exists)
Activity
don-petry commentedon Sep 21, 2026
Blocking analysis — the ring soak and this story are circular (2026-09-21)
Verified against
standards/canary-rings.jsonand the live repo contents.persona-mention ring membership vs. actual adoption:
.github/workflows/persona-mention.yml?.github-private.githubTalkTerm,bmad-bgreat-suite*persona-mention/v1-ring0was advanced tofde249e0on 2026-09-21 (identical tov1-next, releasev1.9.0). That soaked nothing — ring0's only member has no caller stub, so no run exercises the tier. The same is true of ring1.The circularity: a real ring soak needs the stub deployed at each tier → deployment is this story → this story is gated on
v1-stableexisting → per #783 stable is the last tier, granted only after ring0 and ring1 have soaked. Each waits on the other.Why the gate exists anyway:
deploy-standard-workflows.sh(~line 528) callsring_tag_existsand, on a miss, logsrefusing to deploy a non-resolving pin, sets_OVERALL_FAILED=1and continues. So listing this stub beforev1-stableexists does not corrupt anything — ring0/ring1 repos deploy correctly and only stable-tier repos are refused — but the weekly sweep (standards-deploy.yml, Mondays 08:16 UTC) is marked failed every run until the stable tag is cut.Two workable orders (maintainer decision, recorded here rather than chosen unilaterally):
standards-deploy.ymlwithrepo=.github, workflow=persona-mention.ymlto deploy ring0 only, observe real router runs, cutv1-ring1, deploy the two ring1 repos, observe, then cutv1-stableand let the Monday sweep finish the fleet. Honours [promote persona-mention 1/3] Soak through the rings (next → ring0 → ring1 → stable) #783's no-skip rule and produces actual soak evidence. Cost: the weekly sweep reports failure for stable-tier repos during the soak window.v1-ring1andv1-stablenow, then land this story; the sweep deploys fleet-wide with correct per-tier pins in one pass. Cost: the ring soak is nominal — no tier was ever exercised before stable, which is what [promote persona-mention 1/3] Soak through the rings (next → ring0 → ring1 → stable) #783 exists to prevent.(A) is the honest reading of #783. (B) is faster and accepts that the soak was already done at
nextin.github-private— where the router has run continuously, including the #1134 stop-marker behaviour merged 2026-09-20.don-petry commentedon Sep 21, 2026
Decision — option (B), tags first, one-pass deploy (maintainer, 2026-09-21)
Recorded so the sequencing is unambiguous for whoever implements this.
v1-ring1andv1-stableare cut in tier order (no tier skipped, per #783's rule), and this story then lands in one pass: the sweep deploys the fleet with correct per-tier pins and no weekly-sweep failures.What this decision explicitly accepts: the ring soak below
nextis nominal.ring0(.github) andring1(TalkTerm,bmad-bgreat-suite) have nopersona-mention.ymlstub, so no run exercises those tiers before stable. The soak evidence that exists is fromnextin.github-private, where the router has run continuously — including the #1134 stop-marker behaviour merged 2026-09-20 and live onv1-next/v1-ring0(fde249e0).Consequence worth stating plainly: once deployed, the fleet's first real exercise of this router is production. The mitigations are the router's own fail-closed design (an unreadable label set or a failed contract fetch refuses to route, rather than routing unchecked) and the fleet-wide
<id>:hands-offopt-out labels provisioned in combsmith/.github-private#1644.AC #1 remains as written and is satisfiable:
v1-stablewill exist, and no tier will have been skipped in tag order. Do not interpret it as requiring soak telemetry from ring0/ring1 — there is none, by this decision.Implementation is unblocked once
persona-mention/v1-stableresolves. Verify that withring_tag_exists(orgit ls-remote) before starting, exactly as AC #1 says — do not take this comment's word for it.don-petry commentedon Sep 21, 2026
Precondition MET — unblocked 2026-09-21
persona-mention/v1-stablenow resolves. Verified via the refs API:persona-mention/v1-nextfde249e0persona-mention/v1-ring0fde249e0persona-mention/v1-ring1fde249e0persona-mention/v1-stablefde249e0compare v1-stable...v1-next→identical. Tiers were advanced in order (ring0 → ring1 → stable), so AC #1's no-skip requirement holds. Per the maintainer decision recorded above, soak evidence comes fromnext(.github-private); there is none from ring0/ring1 and none is required.Re-verify this yourself with
ring_tag_exists/git ls-remotebefore starting (AC #1 says in code, not from a comment) — then proceed with AC #2 onward. AC #3's dry-run must still show every computed pin resolving before any PR is opened.don-petry commentedon Sep 21, 2026
Dev-Lead Implementation Plan
Issue: #1156 — [qa-lead reach 2/3] Add persona-mention.yml to DEPLOYABLE_WORKFLOWS — blocked on the v1-stable ring tag (#783)
Precondition re-verification (AC #1) — VERIFIED IN CODE ✅
Contrary to the issue's 2026-09-21 snapshot ("there is no
v1-stable"), the ring soak(#783) has since completed. Verified via the live code path the sweep uses:
git ls-remote --tagsonpetry-projects/.githubshowspersona-mention/v1-stablenow exists, pointing at
fde249e0— the same commit asv1-next,v1-ring0,and
v1-ring1. Every tier is aligned at one commit, so stable never got ahead ofring1 → the soak reached stable without skipping a tier.
ring-pins.sh:persona-mentionis ring-managed, channel major =1, astable-tier repo computes
@persona-mention/v1-stable, andring_tag_existson thehost confirms it resolves. Ring members (
v1-next/v1-ring0/v1-ring1) all resolve too.The #1088 assert-exists precondition is therefore satisfied; listing the stub is safe.
Scope
Make
persona-mention.ymleligible for the weekly standards sweep by adding it toDEPLOYABLE_WORKFLOWS— the template and ring registry already exist; this is listing only.Implementation Checklist
scripts/deploy-standard-workflows.sh— addpersona-mention.ymltoDEPLOYABLE_WORKFLOWSuses:pin (AC Add workflow, environment, and orchestration guidance #4); nocheckout_ref:line exists in itreconcile_skip_repo_required_workflowsgap (persona-mention is NOT a required workflow)Tests to Write
test/scripts/deploy-standard-workflows/persona-mention-deployable.bats:is_deployable_workflow persona-mention.ymlis true (list membership, AC Add stacked PR strategy and Epic-level workflow guidance #5)persona-mentionin a stable-tier repo resolves to@persona-mention/v1-stable(AC Add stacked PR strategy and Epic-level workflow guidance #5), via a--dry-runwith a fakeghVerification / AC #3 evidence
--dry-run --workflow persona-mention.ymlagainst the org, output pasted into the PR,showing every target repo pinned to an existing tag (stable →
v1-stable, ring members →v1-<tier>).Test Command
bats test/scripts/deploy-standard-workflows/ test/scripts/lib/ring-pins.batsandbash .dev-lead/scripts/dev-lead-lint.shdon-petry commentedon Sep 21, 2026
Dev-Lead: Progress (in progress — not yet durable)
Precondition (AC #1) — re-verified in code
The 2026-09-21 snapshot said "there is no
v1-stable". That is no longer true — #783completed. Verified via the sweep's own code path (
ring-pins.sh) andgit ls-remote:persona-mention/v1-stableexists atfde249e0, the same commit asv1-next,v1-ring0, andv1-ring1. All tiers aligned at one commit ⇒ stable never got ahead ofring1 ⇒ the soak reached stable without skipping a tier.
ring_tag_existson the hostconfirms the computed stable-tier pin resolves.
Plan Execution (intended)
persona-mention.ymltoDEPLOYABLE_WORKFLOWS(listing only; template + ringregistry entry already existed). Template's
uses:pin left untouched (AC Add workflow, environment, and orchestration guidance #4; nocheckout_ref:line exists in it).v1-stableemission +assert-exists refusal regression (AC Add stacked PR strategy and Epic-level workflow guidance #5). Auto-discovered by the existing Standards
Deploy Tests bats job (AC feat: add Structured Logging and CQRS standards #6).
AC #3 — live
--dry-run --workflow persona-mention.yml(provisional evidence, no PRs opened)Every computed pin resolves to an existing tag (stable-tier →
v1-stable; ring1 members→
v1-ring1). The run exited 0 with no assert-exists refusal — no fleet-breaking pin.Files touched (this workspace, not yet pushed)
scripts/deploy-standard-workflows.sh: one line —persona-mention.ymladded toDEPLOYABLE_WORKFLOWS.test/scripts/deploy-standard-workflows/persona-mention-deployable.bats: new test (3 cases).Notes
persona-mentionis NOT in the audit'sREQUIRED_WORKFLOWS, so noreconcile_skip_repo_required_workflowsgap is introduced (adoption ≠ reach). Confirmed:the "shipped deploy config reconciles…" guard test still passes.
--severity=warning -xclean;.dev-lead/scripts/dev-lead-lint.shall checks passed; the Standards Deploy bats suite is green.standards/pr-limits.json, 50 org-wide):this stub's sweep fans out one PR per stable/ring1 repo.
don-petry commentedon Sep 21, 2026
Dev-Lead: Implementation Complete — PR #1159
The implementation for issue #1156 is durable: commits are pushed and a pull request is open. This record is backed by verifiable artifacts you can confirm in one click.
e80776bd4afb12299726d832d8b2725a92f8efcfReview happens on the PR. Acceptance criteria and test results are evidenced by the PR's diff and CI checks — not asserted here. If a later stage fails, this claim is retracted in place.
feat: implement issue #1156 — [qa-lead reach 2/3] Add persona-mention…