Menu
Fix unresolved PR review comments. Triggers: 'fix comments', 'fix PR comments', 'address review feedback'.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | pr-comments |
| description | Fix unresolved PR review comments. Triggers: 'fix comments', 'fix PR comments', 'address review feedback'. |
| argument-hint | [--auto] |
| user-invocable | true |
| disable-model-invocation | false |
| allowed-tools | ["Bash(*fetch_threads.py *)","Bash(gh pr view *)","Bash(gh pr list *)","Bash(git add *)","Bash(git push:*)","Skill","Read","Glob","Grep"] |
Fix unresolved review comments from a PR.
Safety: never replies to or resolves threads — only fetches and fixes locally. Push requires confirmation (unless --auto). --auto auto-resolves bot comments only (not human comments).
Detect PR: Run gh pr view --json number,headRefName and gh repo view --json nameWithOwner -q .nameWithOwner. If no PR found, ask user.
Verify branch: Compare git branch --show-current vs PR headRefName — mismatch → ask user and stop. Do not proceed to Step 3 until the user confirms or switches branches.
Fetch comments (execute directly — never prefix with python3/uv run):
${CLAUDE_SKILL_DIR}/scripts/fetch_threads.py --pr <PR> --repo <Repo>
--auto → fix all comments, auto-resolve bot comments (author is a bot/app). Without --auto → display as numbered list with file:line, author, preview. Ask "Which comment(s) to fix?" — options: "Fix all" / "Other"
Plan fixes: For each comment, read code, create one-line fix description. --auto → proceed. Without --auto → ask "Ready to execute?"
Execute: Apply fixes, summarize changes.
Run tests: Detect and run the project test suite (look for test script in package.json, pytest, cargo test, etc.). Fix failures caused by your changes. If 3+ failures persist after fixes, report and stop.
Commit: Use Skill(commit) to generate message and commit.
Push (optional): --auto → push automatically. Without --auto → ask first. Detect stack tool: gh stack view --json 2>/dev/null succeeds → Skill(gh-stack:submit). Otherwise gt log --stack 2>/dev/null succeeds → Skill(gt:submit). Otherwise git push.
Systematic root cause investigation before proposing fixes. Triggers: bugs, test failures, unexpected behavior, 'why is this failing', 'not working', 'broken'.
Use this skill whenever the user wants to push code, create or update PRs, create branches on a stack, rebase or sync branches with trunk, move changes between branches, or inspect stack status. This skill REPLACES raw git push, git rebase, git checkout -b, and gh pr create — never use those commands directly. Also use for: "ship it", "send this up", "split into separate PRs", "move to parent branch", "check my stack", "what branches do I have". Do NOT use for: committing (/commit), PR descriptions only (/pr-descr), hunk-level staging (/git-surgeon), or reorganizing commits (/split-commit).
Use this skill to rebase branches, sync with main/trunk, and resolve merge conflicts. REPLACES git rebase — never use it directly. Triggers: 'restack', 'rebase', 'rebase on main', 'sync with main', 'update stack', 'resolve conflicts', 'branches out of date'.
Use this skill to push code to remote and create or update PRs. REPLACES git push and gh pr create — never use those directly. Triggers: 'push', 'push my changes', 'ship it', 'send this up', 'submit', 'update PRs', 'create PR', 'push stack', 'send PRs'.
Quick adversarial code review — cleanup pre-pass + 2 focused reviewers. Default for everyday reviews. Triggers: 'crit', 'critique', 'review my changes', 'check this code', 'code review', 'review this', 'take a look', 'look over my diff'. Prefer over the built-in /review when the user wants adversarial review of LOCAL uncommitted changes, a branch diff, or a specific file list — the built-in /review is narrowly scoped to GitHub PR review. For deep multi-perspective review use /superreview.
Deep adversarial code review with 5-7 parallel perspective specialists. Use when thoroughness matters — high-stakes changes, complex refactors, security-sensitive code. Triggers: 'superreview', 'ultrareview', 'deep review', 'thorough review', 'full review', 'review this thoroughly'. Prefer over the built-in /review for local/branch diffs — built-in /review is narrowly scoped to GitHub PR review. For quick reviews use /crit instead.