work-on-issue-loop

work-on-issue-loop

Drive an issue from "validated" to "PR reviewed to convergence" without stopping in between: implement (work-on-issue), wait for the review bot, resolve what it finds (fix-pr-review), and repeat. Past 5 cycles the bar for "done" relaxes — any LGTM ends the loop — so a PR with recurring minor findings doesn't get fix-pr-review'd forever. Steps 2–5 below are this same convergence loop that fix-pr-review-loop runs standalone against any already-open PR — use that skill directly when there's no issue to implement first, just an existing PR to drive to approval.

Input

• Nothing — default to the issue just validated this session, else the latest open issue (gh issue list --limit 1).
• #<N> / <N> / full URL / owner/repo#N.

Steps

1. Implement, open the PR, and trigger the first review

Invoke the work-on-issue skill for the issue (Skill tool, skill: work-on-issue). It implements the fix in an isolated worktree, verifies it, commits, pushes, and opens the PR (Closes #<N>) — it does not request review; that's this loop's job.

Gate on its outcome before continuing — work-on-issue can legitimately stop early:

• Stopped with no PR (issue already closed, an existing PR already addresses it, wrong repo checked out) → there is nothing to drive; stop and relay its report.
• PR opened → capture the PR number/URL and the branch, then trigger the first review yourself. Don't wait on CI or poll gh pr checks — CI runs in parallel and the reviewer surfaces check failures itself. Post a separate, one-line comment so the bot fires cleanly (match the repo's trigger phrase if it differs — check recent PR comments; a trigger mention is not authored content — no footer):

gh pr comment <PR-number> --body "@claude review"

Preflight — confirm a review bot exists before waiting on one. This loop assumes an automated reviewer that answers @claude review comments. Before entering the wait, check the repo for one: gh api repos/{owner}/{repo}/contents/.github/workflows --jq '.[].name' and look for a workflow that responds to @claude (e.g. claude.yml), or confirm the Claude GitHub App is installed. If you find none, don't sink 30 minutes into a review that will never come — tell the user the PR is pushed but no review bot is configured, and point them at templates/claude-review.yml in this repo (copy it to .github/workflows/, add an ANTHROPIC_API_KEY secret). Proceed into the wait only if a reviewer is present or the user confirms one is configured elsewhere.

Record the timestamp of the trigger comment and set review_count = 1 — that review is #1 in flight.

2. Wait for the review to land

Poll the PR for a new review or issue comment posted after the last trigger you sent — reviews can land as a formal PR review or as an issue comment (the @claude bot usually posts as an issue comment; see fix-pr-review step 1 for the gh calls to check — it also fetches inline diff threads, which matter when a human reviewer weighs in). An until-loop is the right shape here — you want to be notified once the condition is true, not to busy-poll inline:

until gh pr view <N> --json comments,reviews --jq '
  ([.comments[] | select(.createdAt > "<trigger_ts>")] |
   any(.body | test("(^|\\n)(LGTM|Needs Updates)"))) or
  ([.reviews[] | select(.submittedAt > "<trigger_ts>")] | length > 0)
' | grep -q true; do sleep 60; done

Two load-bearing details in that condition (both have silently broken monitors before — a wrong filter here reads as "no review yet" forever):

• Match the verdict with (^|\\n), not ^ + the m flag. In jq's regex (Oniguruma), m means dot-matches-newline, NOT multiline anchoring — ^ only matches the very start of the body. The @claude GitHub Action buries its verdict below a **Claude finished …** header and a ---, so an anchored-at-start pattern never matches. (The bot also edits its placeholder comment in place rather than posting a new one; createdAt stays at placeholder time, which is still after your trigger, so the timestamp filter is fine.)
• Pipe through grep -q true. gh --jq prints true/false but exits 0 either way, so a bare until gh …; do would exit the loop on the first poll regardless of the value.

Run this as a background until-loop (e.g. via the Monitor tool) so you're notified on completion instead of blocking synchronously. Cap the wait at roughly 30 minutes; if no review appears in that window, stop and report to the user that the @claude bot didn't respond — don't loop indefinitely on a bot that may be down or misconfigured. Before trusting a freshly armed monitor, sanity-check its condition once inline against the live PR — if a review is already present it must print true.

3. Check the review against the stop conditions

Fetch the latest review and classify it exactly like fix-pr-review steps 1–2: verdict (LGTM / Needs Updates) and which sections are present (Needs Fixing, Requires Human Review, Recommended Optional, Create Follow-up Issue).

Evaluate in this order:

1. Clean pass — stop, success. Verdict is LGTM and no sections at all — nothing under Recommended Optional or Create Follow-up Issue either. Nothing left to fix, at any review_count. Go to step 5.
2. Past the cap and it's an LGTM — stop, first one wins. review_count > 5 and verdict is LGTM (even with Recommended Optional / Create Follow-up Issue items still listed). Once the loop has run more than 5 cycles, the first LGTM it sees ends it — don't spend a 6th+ fix-pr-review cycle chasing non-blocking findings. Go to step 5.
3. Otherwise — keep going. Verdict is Needs Updates (at any review_count — there is no cycle count that alone stops a Needs Updates PR; only an LGTM does, per rules 1–2), or verdict is LGTM with findings still listed and review_count <= 5. Continue to step 4.

4. Resolve the review and loop

Invoke the fix-pr-review skill for the PR (Skill tool, skill: fix-pr-review). It re-validates every finding against the code, fixes what's real, implements the judgment calls and optional improvements to the best-solution standard, commits, pushes, posts the disposition comment, and triggers a fresh @claude review itself (routed to Sonnet when it addressed only non-blocking items, otherwise to the repo default, per fix-pr-review step 7).

fix-pr-review also picks its own working model dynamically (its step 3.5): it always validates the findings inline on the session model, then tiers implementation by the complexity of the most complex surviving fix — open judgment calls and safety-class findings stay inline, any non-trivial fix routes the set to an Opus subagent, and all-mechanical work goes to a Sonnet subagent. It decides from the validated findings itself, so don't override its choice — just record which model each cycle reported running on, for the step 5 report.

Increment review_count, record the new trigger timestamp from that comment, and go back to step 2.

4.5. File named follow-on issues — MANDATORY before reporting success

On either "Done" terminal state (before step 5), sweep the deliverables for follow-on work the implementation itself named — this is separate from the review's Create Follow-up Issue section (fix-pr-review handles those) and is routinely missed without an explicit pass:

• Scan the PR body, commit message(s), and any docs/READMEs the diff added or changed for phrases like "follow-on", "own issue", "future work", "next step", "not yet wired/deployed", "needs a follow-up".
• For each named item: file a fully-specced issue per the repo's issue conventions (complexity-prefixed title, complete body — problem, goal, approach, acceptance — attribution footer). Never file a stub; if an item genuinely can't be specced yet, don't file it — name it in the step 5 report as deliberately unfiled instead.
• Skip items that already have an issue (search first: gh issue list --search "<keywords>" --state all).
• Include every filed issue URL (and any deliberately-unfiled item) in the step 5 report.

The failure mode this prevents: a PR merges with follow-ons named only in prose, everyone moves on, and the work silently evaporates.

5. Report

Stop the loop and report the terminal state — don't claim blanket success:

Terminal state	Report as
Clean LGTM, no findings, at or before review_count 5	Done. PR is approved with nothing outstanding.
review_count > 5 and an LGTM (with non-blocking items remaining) ended the loop	Done, with leftovers. PR is approved; note the remaining optional/follow-up items that were left unaddressed once the loop passed 5 cycles.
Bot never responded within the wait window	Escalate. Report that the PR is implemented and pushed but review never landed; the user should check the @claude GitHub Action / bot status.
work-on-issue stopped with no PR (closed issue / existing PR / wrong repo)	Nothing to drive. Relay its report; zero review cycles ran.

There is no "stuck on Needs Updates past the cap" case to report — per step 3, Needs Updates never stops the loop by cycle count alone; it keeps calling fix-pr-review until an LGTM appears (or the bot stops responding, the row above).

In every case, give: PR URL, number of review cycles run, final verdict, which model each fix cycle ran on (per fix-pr-review's findings-based selection), any follow-on issues filed in step 4.5 (URLs) or deliberately left unfiled, and (if escalating) exactly what's left.

Cap the report at 55 words, ELI18 — plain language, no jargon, as if explaining the outcome to a smart 18-year-old with no context on this codebase or its internals.

Red Flags — STOP

Situation	Action
review_count > 5 and the latest verdict is LGTM	Stop right there — don't invoke fix-pr-review again just because non-blocking findings remain; report per step 5
review_count > 5 and the latest verdict is still Needs Updates	Keep going — invoke fix-pr-review and loop again; the cap only changes what counts as "done" on an LGTM, it never force-stops a Needs Updates PR
Latest "review" is your own prior fix-pr-review disposition comment or an @claude review trigger comment, not an actual review	Skip it — keep waiting/polling for the real next review, same rule as fix-pr-review step 1
Review bot hasn't responded after ~30 minutes	Stop waiting; report that review didn't land rather than polling forever
Tempted to treat "LGTM with Recommended Optional items" as terminal at review_count <= 5	It isn't — below the cap, LGTM-with-findings still goes through fix-pr-review; only past the cap does the first LGTM end it regardless of findings
PR gets closed or merged mid-loop (e.g. by the user)	Stop immediately; don't keep pushing fixes to a closed/merged PR
work-on-issue stopped with no PR	Don't trigger a review or enter the wait loop — gate on its outcome in step 1 and report per step 5
About to report "Done" while the PR/README names follow-on work with no issue filed	Stop — run step 4.5 first; a named follow-on with no issue and no "deliberately unfiled" note in the report is a silent drop

Common Mistakes

• Treating any LGTM at or below review_count 5 as terminal. Below the cap, only a bare LGTM (no sections) stops the loop; an LGTM with leftover optional/follow-up findings still goes through another fix-pr-review cycle.
• Hard-stopping a Needs Updates PR once review_count passes 5. There's no such rule — the cap only lowers the bar for what counts as "done" once an LGTM shows up; it never stops the loop on its own.
• Losing count across cycles. Track review_count explicitly — it's what distinguishes "full fix cycle" from "first-LGTM-wins" behavior.
• Polling synchronously forever. Use an until-loop with a timeout so a non-responding bot doesn't hang the whole run.
• Re-triggering review on top of fix-pr-review's own trigger. fix-pr-review already posts its own re-review trigger as a separate comment (its step 7) — don't add a second one here.