Repository navigation
(standardize-repo): Provision the model labels before the draft PR instead of after merge #1531
Description
Problem
A standardize-repo update cannot reach its readiness gate until the repository's live labels include the model labels the implementer record writes. The skill provisions those labels only after the update PR merges.
implement/assets/record-implementer.shlabels the issue and the PR with exactly onemodel-family:*,model:*andmodel-version:*label, and at most onemodel-effort:*label.- The integrate stage re-checks that record before the gate (harmon-devkit#1262).
- A template update that adds a model line, such as
model:solormodel-version:6.1, ships the label-registry entry in files. The live label does not exist yet. references/mode-update.md§6 runstask setup:github-labelsonly once the update PR has merged. It gives a reason: that task isgh label create --force, which overwrites the color and description of existing labels.
The writer therefore refuses, and the lane stops to ask. In the fleet rollout every wave-2 and wave-3 lane stopped on this. The orchestrator ran task setup:github-labels by hand in each repository, for example ponderous-content (221 ok), lawnomator-site (218 ok), omator (227 ok) and harmon-infra (222 ok).
Creating a label that does not exist yet is not the hazard the post-merge rule guards against. Overwriting an existing one is.
Current violation (observed 2026-10-17)
references/mode-update.md §6 gates every metadata task, labels included, on the merged PR. No step before the draft PR provisions the model labels the implementer record needs.
Acceptance criteria
- [CI] The label setup has an additive mode that creates only missing labels and never changes an existing label's color or description. A test pins that an existing label is left byte-identical.
- [CI]
mode-update.mdadds a step before the draft PR. It runs that additive mode for the target release's label vocabulary, after the maintainer approves it, and records the result. - [CI] The step accepts one approval covering a whole fleet run, so each lane does not ask again.
- [CI] §6 keeps the post-merge
--forcereconciliation unchanged, and says the additive step does not replace it.
Verify
grep -n 'Run this only once the update PR has merged' ai/skills/repo/standardize-repo/references/mode-update.md
grep -n 'setup:github-labels' ai/skills/repo/standardize-repo/references/mode-update.mdWhile this is open, the first command matches in §6, and every setup:github-labels hit is in §6 or later.
Out of scope
Retired-label migration and the project and issue-field reconciliation, which stay post-merge.
Provenance
Retro of the session on 2026-10-17; the PR and its linked issues carried no Dev flow v2 evidence marker. The run was the 2026-10-13/16 standardize-repo fleet rollout of harmon-init v5.7.2 to v5.8.2 across 13 repositories. Among its PRs were harmonops/harmon-infra#862, sommerlawn/sommerlawn-site#527 and ponderousdev/lawnomator-site#255.
Activity
evanharmon1 commentedon Nov 2, 2026
Claiming — starting implementation on branch feat/1531-additive-model-labels (session fleet-m6-foundations).
Preflight (§3):
Findings:
Claim record (for
/wrap— undo only what this claim added):claim:label added by this claim: claim:gptclaim:model label added by this claim: n/aclaim:label displaced by this claim: noneclaim:label owned by this claim chain: claim:gptclaim:model label owned by this claim chain: n/aclaim:label displaced by this claim chain: noneevanharmon1 commentedon Nov 2, 2026
Slate membership:
evanharmon1 commentedon Nov 3, 2026
Dev flow v2 evidence: challenge round 1 for
run-1531-additive-model-labelsat9e0d274(findercodex-adversarial).{ "run_id": "run-1531-additive-model-labels", "stage": "challenge", "round": 1, "reviewed_head": "9e0d2740f87c0ce9c39a3fecea57284deae75e5c", "passes": [ { "schema": 2, "role": "challenger", "status": "completed", "head": "9e0d2740f87c0ce9c39a3fecea57284deae75e5c", "produced_at": "2026-11-03T13:59:18Z", "producer": { "harness": "codex-cli", "model": "gpt-6-astra", "tier": "apex" }, "run": { "run_id": "run-1531-additive-model-labels", "initiated_by": "human" }, "payload": { "stage": "challenge", "round": 1, "reviewed_head": "9e0d2740f87c0ce9c39a3fecea57284deae75e5c", "finder": "codex-adversarial", "slot": "codex-adversarial", "findings": [ { "id": "challenge-r1-codex-adversarial-1", "path": "ai/skills/repo/standardize-repo/SKILL.md", "line": 79, "class": "consistency", "provenance": "original", "fingerprint": "new", "priority": "P2", "recommended_disposition": "fix", "evidence": "Both `docs/decisions/2026-10-26-fleet-standardization-runs.md` and `specs/standardize-repo-fleet.md` (§ Start-of-run questions) state that `SKILL.md` lists the ordered question set with each question's default, its decisions-file key, and the later stop it removes. `ai/skills/repo/standardize-repo/assets/fleet-decisions.py` adds `label_provisioning` as the 6th question in `QUESTIONS` (shifting the question count from 10 to 11), and `references/fleet-decisions.md` + `references/mode-update.md` document it, but `SKILL.md` lines 69-79 still list only the previous 9 rows in the ordered question table and omit `label_provisioning`." } ], "counts": { "P0": 0, "P1": 0, "P2": 1, "P3": 0 }, "attack_scenarios": [ { "id": "as-1", "description": "Case-insensitive collision & existing-label preservation: verify that when a live label exists with different casing (`Model:Sol`) and custom color/description, `ensure-missing-labels.sh` matches it case-insensitively (`ascii_downcase`), records it in `already_present`, and never invokes `gh label create --force`, `gh label edit`, or `gh label delete`.", "outcome": "held", "finding_id": null }, { "id": "as-2", "description": "Truncated or malformed live label listing: verify that `ensure-missing-labels.sh` requests `--limit 10000` and refuses before any write when `gh label list` fails, returns non-array JSON, or returns `length == 10000` (which could hide a subsequent page and cause an existing label to be treated as missing).", "outcome": "held", "finding_id": null }, { "id": "as-3", "description": "Missing or failing target renderer / malformed renderer output: verify that `ensure-missing-labels.sh` fails closed before listing or creating labels if `$checkout/scripts/label-registry-render.mjs` is missing, exits non-zero, emits records violating `^(?<name>[^|]+)\\|(?<color>[a-fA-F0-9]{6})\\|(?<description>.*)$`, or contains control characters or case-insensitive duplicate names.", "outcome": "held", "finding_id": null }, { "id": "as-4", "description": "Mid-flight `gh label create` failure & partial progress accounting: verify that if a label creation fails after earlier labels were created, `ensure-missing-labels.sh` halts immediately without attempting rollback/deletion, emits the already-created names in `created` alongside the failed label in `refused`, and exits 1.", "outcome": "held", "finding_id": null }, { "id": "as-5", "description": "Bash 3.2 `set -u` portability and profile flag forwarding: verify that `renderer_args=(labels)` is never empty under `nounset`, and optional `--foreman` and `--release-please` flags are forwarded cleanly to `label-registry-render.mjs`.", "outcome": "held", "finding_id": null }, { "id": "as-6", "description": "Fleet approval gating and start-of-run question table synchronization: verify that `label_provisioning` defaults to `withheld` (never implicit approval), is validated against `$defs.label_provisioning` (`approved` | `withheld`), is reusable across repositories within the same release in `fleet-decisions.py`, and is documented consistently across `SKILL.md`, `references/fleet-decisions.md`, and `references/mode-update.md`.", "outcome": "surfaced-finding", "finding_id": "challenge-r1-codex-adversarial-1" } ] } } ], "adjudication": { "schema": 2, "run_id": "run-1531-additive-model-labels", "stage": "challenge", "round": 1, "reviewed_head": "9e0d2740f87c0ce9c39a3fecea57284deae75e5c", "adjudications": [ { "finding_id": "challenge-r1-codex-adversarial-1", "reviewer_priority": "P2", "adjudicated_priority": "P2", "disposition": "defer", "classification": "confirmed", "override": null, "reason": "Confirmed: ai/skills/repo/standardize-repo/SKILL.md lines 69-79 list the 9 start-of-run questions without label_provisioning, while fleet-decisions.py, references/fleet-decisions.md, and references/mode-update.md include it. Deferred to integration because SKILL.md is owned by wave-1 lane-1670 (#1670) and outside lane-1531's fence; will be settled at integration by filing a follow-up or after #1670 merges.", "evidence": "Verified in ai/skills/repo/standardize-repo/SKILL.md:69-79 and plan.json revision 233 (SKILL.md is fenced to lane-1670, not lane-1531)." } ] } }evanharmon1 commentedon Nov 3, 2026
Dev flow v2 evidence: review round 1 for
run-1531-additive-model-labelsat9e0d274(findercodex-verification).{ "run_id": "run-1531-additive-model-labels", "stage": "review", "round": 1, "reviewed_head": "9e0d2740f87c0ce9c39a3fecea57284deae75e5c", "passes": [ { "schema": 2, "role": "reviewer", "status": "completed", "head": "9e0d2740f87c0ce9c39a3fecea57284deae75e5c", "produced_at": "2026-11-03T14:10:26Z", "producer": { "harness": "codex-cli", "model": "gpt-6.1-sol", "tier": "frontier" }, "run": { "run_id": "run-1531-additive-model-labels", "initiated_by": "human" }, "payload": { "stage": "review", "round": 1, "reviewed_head": "9e0d2740f87c0ce9c39a3fecea57284deae75e5c", "finder": "codex-verification", "slot": "codex-verification", "findings": [], "counts": { "P0": 0, "P1": 0, "P2": 0, "P3": 0 } } } ], "adjudication": { "schema": 2, "run_id": "run-1531-additive-model-labels", "stage": "review", "round": 1, "reviewed_head": "9e0d2740f87c0ce9c39a3fecea57284deae75e5c", "adjudications": [] } }github-actions commentedon Nov 3, 2026
Claim released — issue closed (completed). (Supersedes the claim record above.)
Released by claim-release automation:
claim:gptlabel: removedevanharmon1: removedevanharmon1 commentedon Nov 3, 2026
Slate membership:
evanharmon1 commentedon Nov 3, 2026
Slate membership: