ci(town-crier): auto-resolve bus thread on 2-approval consensus

[jasperboerhof]

Adds the 2-approval consensus auto-resolve job to this repo's town-crier producer workflow, matching the version merged to kendo (development, 2026-06-23).

What it does: on an approved review of an Agent Review Requested PR, it tallies distinct current-head approvals; once ≥2 agree with no outstanding CHANGES_REQUESTED it POSTs /resolve to the bus to de-announce the thread (further review turns are churn once two independent approvals agree). Stale approvals cast on an earlier head are filtered out, so a regression pushed onto a reviewed head can't satisfy consensus.

Also adds the synchronize re-announce trigger (re-review on push) — this repo's workflow was missing it, bringing it to parity with the other producers.

No new secrets — reuses the existing TOWN_CRIER_URL variable + TOWN_CRIER_TOKEN secret; the consensus job only adds in-workflow pull-requests: read.

The announce-pr.yml is byte-identical to kendo's deployed version (sha256 eb9ccfe3…).

🤖 Generated with Claude Code

[jasperboerhof]

Town Crier Review · 9/10 · PASS · 🔎 Independent

kendo-report-tool #7 · AC anchor: PR description · head b2fc049fa1 · via the town-crier bus (request #157)

Tip

Single-file CI-config PR adding a 2-approval consensus auto-resolve job (plus a synchronize re-announce trigger) to this repo's town-crier producer; I independently simulated the consensus jq across the off-happy-path cases (stale CHANGES_REQUESTED, dismissed/pending reviews, empty review set, approve-then-CR) and verified the byte-identical-to-kendo claim, finding the head-filtered approval tally, the fail-open-but-loud-on-missing-provisioning policy, and the job-scoped pull-requests: read all correct. Clean parity port of an already-deployed workflow with no defect.

No findings — clean against the review checklist.

[Goosterhof]

✅ Approve-worthy

0 blockers · 0 concerns · 1 nit · 1 praise · 0 inline

Round 1. Bus thread round 2 — confirming @Jasper's independent PASS 9/10.

Single-file CI-config PR porting the 2-approval consensus auto-resolve job (plus the synchronize re-announce trigger) into this repo's town-crier producer, bringing it to parity with kendo's deployed announce-pr.yml. The byte-identity claim checks out: the head file hashes to eb9ccfe3b280577ae8a9f08c4afee134da8e9185d56e8cb1b97118636f3e3074, identical to origin/development:.github/workflows/announce-pr.yml on kendo. CI green (announce, check (8.4), check (8.5)); consensus/resolve skip correctly on this event.

Cross-file findings

Head-filter asymmetry (nit, not a defect). The consensus tally filters approvals to the current head (.state == "APPROVED" and .commit_id == $head) but counts CHANGES_REQUESTED head-blind. I walked the two regimes: with dismiss_stale_reviews on, a stale CR is DISMISSED and the latest-per-reviewer group_by | max_by(.submitted_at) drops it; with the setting off, a CR cast on an earlier head stays CHANGES_REQUESTED and blocks consensus even after the concern is fixed on a new head. That is the conservative direction — a stale CR holds the thread open rather than waving a regression through — so it is correct as written. Worth a one-line comment that the asymmetry is deliberate, since a future reader will ask why one side is head-filtered and the other is not.

Praise: the NDJSON-via---paginate --jq '.[]'-then-jq -s slurp explicitly avoids gh api --slurp (newer-gh only) — a real portability decision against the runner's gh version, not boilerplate, and the inline comment naming it earns its place.

Verified parity against kendo and re-simulated the head-filter edge that @Jasper already covered; nothing to add to the consensus correctness case. Clean port.

Automated war-room agent review — posted because this PR carries the Agent Review Requested label.

The merge-base changed after approval.

[jasperboerhof]

Town Crier Review · 9/10 · PASS · 🤝 Confirm

kendo-report-tool #7 · AC anchor: PR description (config-only parity port of kendo's announce-pr.yml 2-approval consensus auto-resolve job; no board issue) · head 2244e78cb9 · via the town-crier bus (request #157)

Tip

Single-file CI-config PR porting kendo's deployed 2-approval consensus auto-resolve job (plus the synchronize re-announce trigger) into this repo's town-crier producer. Corroborates the prior reviews: byte-identity to kendo's origin/development verified independently (sha256 eb9ccfe3…), and the consensus jq holds across the off-happy-path cases — head-filtered approval tally, latest-per-reviewer dedup, and job-scoped pull-requests:read least-privilege all check out. The one nit raised on the thread (CHANGES_REQUESTED tally is head-blind, unlike the head-filtered approvals) is benign by direction — a stale CR can only block consensus, never cause a false auto-resolve. Clean parity port; nothing refuted.

No findings — clean against the review checklist.

Bus thread · 2 prior review(s):

• jasper (independent): Independent first look (empty thread). PASS 9/10 — no findings. Single-file CI-config PR adding a 2-approval consensus a
• general (confirm): Approve-worthy. Confirm @Jasper's PASS — byte-identity to kendo verified (sha256 eb9ccfe3…), CI green. One nit: CHANGES_

[jasperboerhof]

Town Crier Review · 9/10 · PASS · 🔎 Independent

kendo-report-tool #7 · AC anchor: PR description (task, no linked issue) · head 2244e78cb9 · via the town-crier bus (request #110)

Tip

Single-file CI-config PR adding a consensus job to announce-pr.yml that retires a bus thread once two distinct current-head approvals agree with no outstanding change request; I traced the new job's trigger gate, jq tally, stale-approval guard, dismissed-review handling, off-happy-path failure modes, and job-scoped permissions. The logic is robust and the failure policy fails loud on missing provisioning yet fails safe (thread stays open, never a wrong resolve) on transients — no defect found.

No findings — clean against the review checklist.

[jasperboerhof]

removed the file