Repository navigation
[SCR-183] Roll the issue standard to the fleet - #197
Conversation
No repo but content-rabbit has issue forms, and 85 open issues across the four active repos carry no labels at all. Plan a repo: the four forms, config.yml, the 13 labels, the AGENTS.md marker block and a .agents/issues.md stub. Dry by default like pr-standards-rollout, so --apply is the only way to write anything. Three reasons to skip, each said out loud rather than passed over in silence. Issues disabled means there is nothing to standardise. An existing .github/ISSUE_TEMPLATE means the repo already adopted it. No .github/pr-standards.json means the branch name cannot be derived, because the issue standard borrows that prefix. Labels are the only part applied directly. Creation is idempotent with --force and a label is not a code change needing review. Everything else goes through an issue and a pull request, because the standard being rolled out says so, and a rollout that exempted itself would be the first thing to argue with. Refuse to start if any template is missing or empty. set -u does not catch a failed read, so an unreadable template would sail past and --apply would write an empty form into every target repo. Stopping before the first remote write beats repairing sixty repos after.
|
Warning Review limit reachedNext included review available in 24 minutes. View limit detailsLimit details: You’ve used the included review currently available. Your 106 included PR review attempts over the past 7 days set your current allowance at 1 review per hour. Your organization has reached its usage spending cap. Adjust your spending cap in the billing tab. Review configuration: ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Team Run ID: 📒 Files selected for processing (1)
Comment |
There was a problem hiding this comment.
Reviewed the diff, council-findings.md, and the PR body/thread against the actual script. No Critical issues. I'm not pushing any fix this cycle (pushing is disabled for this run) — flagging three confirmed Major items and two Minor items for a human.
Confirmed, open
-
Verification evidence isn't reproducible from this diff.
REQUIRED_TEMPLATES(lines 43-47) requiresissue-standards-templates/*, which doesn't exist in this diff and isn't onmain— those files live in the still-open #196. Checking out exactly this branch and running./issue-standards-rolloutexits 2 before any planning happens, yet "How I verified" pastes a full successful 133-repo dry run. That output could only have come from an environment with #196's templates also present locally — it doesn't demonstrate what ships in this PR by itself. Settling this needs either a re-run in a clean checkout of just this branch (showing the expected exit 2), or an explicit note that verification depends on #196 landing first. -
The only side-effecting path (
--apply) is completely unverified. Lines 145-173 are new: the label-creation loop, per-label failure detection,label_failurestally, and the exit-1 branch. The PR body says outright "--applyhas not been run against anything." That's the one code path here with real, remote, side effects, and it has zero command-result evidence. -
repo_prefix()still conflates "file absent" with "read failed." Lines 91-100: a rate-limited or malformedpr-standards.jsonread is caught by the blanketexcept Exceptionand prints'', whichplan_repothen reports asSKIP ... no .github/pr-standards.json— a misleading reason for a transient API failure. Notably, the immediately preceding commit (f37bdf0) hardened the sibling functionrepo_already_adopted()against this exact ambiguity with a three-way yes/no/unknown result, butrepo_prefixwasn't given the same treatment, so the gap is still open right next to its own fix.
Minor (not blocking)
-
N+1 API calls for
has_issues.plan_repocallsgh api repos/$1per repo (line 71) purely to readhas_issues, even thoughtarget_repos()already does onegh repo list --jsoncall for the whole fleet (line 107) — that endpoint can returnhasIssuesEnableddirectly. Across the fleet run (133 repos per the PR's own numbers) this is ~133 avoidable serial round-trips. -
"GitHub always creates
bug" (lines 49-51) is an assumption, not a guarantee. Repos created via API without default labels, or repos wherebugwas later deleted, would silently lack that type label since it's intentionally excluded from the managed set. Worth a spot-check during rollout, not a blocker.
Discarded
- Council flagged "the script never opens the issue/PR, only creates labels" (both correctness and scope lenses). This is explicit, stated intent, not a bug: the script's own comments (147-150, 171-173) and the PR body ("Everything else goes through an issue and a pull request, because the standard being rolled out says so") both say the stub can't be filled from outside the repo, so that step is deliberately left to a human.
PR-standard lens
Branch, title, and body all match .github/pr-standards.json (prefix scr): branch scr-183-..., title [SCR-183] Roll the issue standard to the fleet, body has exactly one Closes #183, ## What/## Why/## How I verified, and ends in Assisted-by:. 196 counted lines, 1 file — well under the 500/40 caps.
🧑⚖️ Council
| Model | Lens | Result |
|---|---|---|
| GPT-5.6 (Codex) | correctness | 1 confirmed (repo_prefix ambiguity), 1 discarded (issue/PR creation — by design), 1 not confirmed (bug-label assumption, noted as minor) |
| Gemini 3 Pro | performance | confirmed (N+1 has_issues calls) |
| Kimi K3 | security | no findings |
| Grok 4.5 | maintainability | timed out — no findings produced |
| GPT-5.6 (scope) | scope | 1 confirmed (templates missing / evidence not reproducible), 1 discarded (issue/PR creation — by design), 1 confirmed (--apply path unverified) |
Closes #183
What
issue-standards-rollout: plans the adoption of the issue standard across the fleet. Dry by default,--applyto write,--repofor one repo.Why
No repo but
content-rabbithas issue forms, and 85 open issues across the four active repos carry no labels at all.Labels are the only part applied directly. Creation is idempotent with
--force, and a label is not a code change needing review. Everything else goes through an issue and a pull request, because the standard being rolled out says so, and a rollout that exempted itself would be the first thing to argue with.It refuses to start if a template is missing or empty.
set -udoes not catch a failed read, so an unreadable template would sail past and--applywould write an empty form into every target repo. Stopping before the first remote write beats repairing sixty repos afterwards. That failure mode is borrowed frompr-standards-rollout, which learned it the same way.How I verified
A full fleet dry run, which wrote nothing:
Skip reasons, which are the interesting half:
Single-repo dry run, and a check that it really touched nothing:
The prefix is read from each repo's own config rather than guessed, which is why
skillsplansski-and not something derived here.What the dry run found
55 of 133 repos have no
.github/pr-standards.json. The issue standard borrows that prefix for its branch name, so the issue rollout is gated behind the PR rollout for those repos. They are skipped with that reason rather than given a guessed prefix, because a guessed prefix produces a branch name the PR checker will then refuse.That is worth knowing before running
--apply: the reachable set today is 58 repos, not 133, and growing it means runningpr-standards-rolloutfirst.The 7 that already have templates are forks that arrived with them. Skipping them is correct, and it is also what makes a rerun a no-op.
Not in this PR
The templates themselves are #196. The two together are 614 counted lines against a 500 cap, and the cap forced the split, which is what it is for.
--applyhas not been run against anything. The plan is readable and the reachable set is known; running it is a separate decision.Assisted-by: claude-personal-1:claude-opus-5