From work-cycle
Use when a PR is open and needs review (including the shorthand "fixpr"). Fetches PR state, reviews the diff, fixes issues directly in the worktree, pushes, and comments on the PR about what was fixed. Can loop until clean.
How this skill is triggered — by the user, by Claude, or both
Slash command
/work-cycle:pr-review-fixThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Review an open PR, fix issues directly, push, and comment. Default behavior is fix-and-document, not report-and-wait.
Review an open PR, fix issues directly, push, and comment. Default behavior is fix-and-document, not report-and-wait.
Uses the same definitions as plan-do-review-renew:
sync — sync local repo main to remote main (git checkout main && git pull). If continuing in a worktree, also sync the worktree's local main reference.
mergesync — merge the PR (gh pr merge --merge --delete-branch), then sync. One atomic operation.
fixpr — shorthand for this workflow. Treat user phrases like fixpr, fix PR, review-fix-comment, "fix this PR and comment", or "review this PR, push fixes, and leave comments" as invocations of this skill. Default mode is fix, not comment-only — switch to comment-only only when the user explicitly says "just review", "comment only", "don't push", or no safe worktree/branch is available.
check — fetch ALL PR state:
gh pr view <number>
gh api repos/{owner}/{repo}/pulls/{number}/reviews
gh api repos/{owner}/{repo}/pulls/{number}/comments
gh pr view <number> --json commits
gh pr checks <number>
Accept PR number, URL, or infer from current branch. Then:
Worktree: <path> in PR descriptions)git worktree list for a worktree on the PR's head branchConfirm before proceeding:
Reviewing PR #
<number>:<title>Worktree:<path>| Branch:<branch>| Mode: fix (or comment-only if no worktree)
User can also explicitly request comment-only mode (e.g. "just review", "comment only", "don't fix").
Read the full diff against the PR's base branch:
Fix mode:
cd <worktree-path>
git diff <base-branch>...HEAD
Comment-only mode (no worktree):
gh pr diff <number>
Review for:
CLAUDE.md for patterns)Classify each finding by severity:
| Severity | Meaning | Action |
|---|---|---|
| Critical | Broken logic, security vulnerability, data loss risk | Must fix |
| Important | Correctness issue, missing edge case, test gap | Fix by default |
| Minor | Style, naming, minor simplification | Fix if trivial, otherwise note |
| Observation | Not wrong, but worth knowing | Comment only, do not fix |
If in comment-only mode, skip to Step 4. After posting the comment, the review is done — do not proceed to Step 5.
Otherwise, proceed to Step 3 (fix is the default when a worktree is available).
For each Critical and Important finding (and trivial Minor findings):
If a fix is non-trivial or ambiguous, ask the user before applying it.
After all fixes:
cd <worktree-path>
git add <fixed-files>
git commit -m "review: fix <concise summary>"
git push
Post a single review comment summarizing all findings and fixes:
gh pr comment <number> --body-file /tmp/review-comment.md
Comment structure:
## Review Summary
### Fixed
- [Critical] <description> — fixed in <commit-sha-short>
- [Important] <description> — fixed in <commit-sha-short>
- [Minor] <description> — fixed in <commit-sha-short>
### Not Fixed (needs discussion)
- [Minor] <description> — <reason skipped>
### Observations
- <observation>
Use --body-file to avoid shell quoting issues with markdown.
Comment-only mode ends here. Do not proceed to Step 5.
After pushing fixes, re-read the diff to check:
If clean (no new issues, no deferred Critical/Important findings), present options:
mergesync — merge the PR now, sync (only offer when invoked standalone, not from within plan-do-review-renew)done — leave PR open, report what was doneanother-pass — run Steps 2–5 again (e.g. if fixes were complex)If new issues found, loop back to Step 3 automatically (max 3 cycles, then present findings to user).
npx claudepluginhub anaka/work-cycle --plugin work-cycleGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.