/review-pr
Review GitHub pull requests with structured code analysis. Use when asked to review a PR, check a pull request, or audit code changes.
$ npx -y skills add suitedaces/dorabot --skill review-pr --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
- Slash command
/review-pr
Context preview
The summary Claude sees to decide when to auto-load this skill.
Review GitHub pull requests with structured code analysis. Use when asked to review a PR, check a pull request, or audit code changes.
SKILL.md
review-pr.SKILL.mdname: review-pr
description: "Review GitHub pull requests with structured code analysis. Use when asked to review a PR, check a pull request, or audit code changes."
metadata:
{ "requires": { "bins": ["gh", "jq"] } }PR Review Skill
Review GitHub pull requests with structured analysis covering correctness, security, performance, and style.
Invocation
Triggered by: `/review-pr <url-or-number>`, "review this PR", "review PR #123", or a GitHub PR URL.
Parse the input:
- Full URL: `https://github.com/owner/repo/pull/123` -> extract owner, repo, number
- Number only: `#123` or `123` -> use current repo from `gh repo view --json nameWithOwner -q .nameWithOwner`
- If ambiguous, ask for the repo.
Step 1: Gather Context
Run these in parallel:
# PR metadata
gh pr view <number> --repo <owner/repo> --json title,body,author,baseRefName,headRefName,createdAt,additions,deletions,changedFiles,labels,reviewRequests,mergeable,state
# Full diff
gh pr diff <number> --repo <owner/repo>
# File list with stats
gh pr view <number> --repo <owner/repo> --json files --jq '.files[] | "\(.additions)+/\(.deletions)- \(.path)"'
# CI status
gh pr checks <number> --repo <owner/repo> 2>/dev/null || echo "No checks"
# Existing review comments (avoid duplicating feedback)
gh api repos/<owner>/<repo>/pulls/<number>/comments --jq '.[] | "[\(.path):\(.line // .original_line)] @\(.user.login): \(.body)"' 2>/dev/null | head -50
Step 2: Analyze the Diff
Read the full diff carefully. For each file, evaluate:
Correctness
- Logic errors, off-by-one, null/undefined handling
- Missing error handling or edge cases
- Broken contracts (function signatures changed without updating callers)
- Race conditions or concurrency issues
Security
- User input flowing to SQL, shell commands, file paths, or HTML without sanitization
- Hardcoded secrets, API keys, or credentials
- Overly broad permissions or missing auth checks
- Unsafe deserialization or eval usage
Performance
- O(n^2) or worse algorithms where O(n) is possible
- Unnecessary re-renders in React (missing memo/useMemo/useCallback)
- N+1 queries or unbounded data fetching
- Large allocations in hot paths
Architecture
- Does this change belong in this file/module?
- Are abstractions appropriate (not too much, not too little)?
- Does it follow existing patterns in the codebase?
- Will this be maintainable?
Style
- Naming clarity
- Dead code or TODOs left behind
- Missing types (in TypeScript)
- Inconsistent formatting (only flag if egregious)
Step 3: Write the Review
Output a structured review in this format:
## PR Review: <title>
**Summary**: 1-3 sentence overview of what this PR does and whether it's ready.
**Verdict**: APPROVE | REQUEST_CHANGES | COMMENT
### Findings
#### <severity-emoji> <category>: <short description>
**File**: `path/to/file.ts` L<line>-L<line>
<explanation of the issue and why it matters>
**Suggested fix**:
\`\`\`diff
- old code
+ new code
\`\`\`
---
(repeat for each finding)
Severity emojis:
- `[blocker]` - Must fix before merge. Bugs, security issues, data loss risks.
- `[warning]` - Should fix. Performance problems, poor patterns, missing edge cases.
- `[nit]` - Optional. Style, naming, minor improvements.
- `[praise]` - Good stuff. Call out well-written code (do this at least once).
Guidelines
- Be specific. Quote the code, give line numbers, explain why.
- Suggest fixes, not just problems. Include diff snippets.
- Don't nitpick formatting if there's a formatter configured.
- If the PR is large (>500 lines), focus on the riskiest files first.
- If the PR description explains a deliberate tradeoff, don't re-litigate it.
- Check the test coverage: are new code paths tested? Are edge cases covered?
Step 4: Post the Review (Optional)
Only post to GitHub if the user explicitly asks. Use:
# Post a review comment (not individual line comments)
gh pr review <number> --repo <owner/repo> --comment --body "<review body>"
# Or approve/request changes
gh pr review <number> --repo <owner/repo> --approve --body "<review body>"
gh pr review <number> --repo <owner/repo> --request-changes --body "<review body>"
For inline comments on specific lines:
gh api repos/<owner>/<repo>/pulls/<number>/comments \
-f body="<comment>" \
-f path="<file>" \
-f commit_id="$(gh pr view <number> --repo <owner/repo> --json headRefOid -q .headRefOid)" \
-F line=<line_number> \
-f side="RIGHT"
Always show the review to the user first and get confirmation before posting to GitHub.
Quick Mode
For small PRs (<100 lines changed), skip the full structure. Give a concise paragraph covering the key points and verdict.
Batch Review
When reviewing multiple PRs (e.g., "review all open PRs"):
gh pr list --repo <owner/repo> --json number,title,author,additions,deletions --jq '.[] | "#\(.number) \(.title) by @\(.author.login) (+\(.additions)/-\(.deletions))"'
Triage by size and risk, review the largest/riskiest first.
Read more
name: review-pr
description: "Review GitHub pull requests with structured code analysis. Use when asked to review a PR, check a pull request, or audit code changes."
metadata:
{ "requires": { "bins": ["gh", "jq"] } }PR Review Skill
Review GitHub pull requests with structured analysis covering correctness, security, performance, and style.
Invocation
Triggered by: `/review-pr <url-or-number>`, "review this PR", "review PR #123", or a GitHub PR URL.
Parse the input:
- Full URL: `https://github.com/owner/repo/pull/123` -> extract owner, repo, number
- Number only: `#123` or `123` -> use current repo from `gh repo view --json nameWithOwner -q .nameWithOwner`
- If ambiguous, ask for the repo.
Step 1: Gather Context
Run these in parallel:
# PR metadata gh pr view <number> --repo <owner/repo> --json title,body,author,baseRefName,headRefName,createdAt,additions,deletions,changedFiles,labels,reviewRequests,mergeable,state # Full diff gh pr diff <number> --repo <owner/repo> # File list with stats gh pr view <number> --repo <owner/repo> --json files --jq '.files[] | "\(.additions)+/\(.deletions)- \(.path)"' # CI status gh pr checks <number> --repo <owner/repo> 2>/dev/null || echo "No checks" # Existing review comments (avoid duplicating feedback) gh api repos/<owner>/<repo>/pulls/<number>/comments --jq '.[] | "[\(.path):\(.line // .original_line)] @\(.user.login): \(.body)"' 2>/dev/null | head -50
Step 2: Analyze the Diff
Read the full diff carefully. For each file, evaluate:
Correctness
- Logic errors, off-by-one, null/undefined handling
- Missing error handling or edge cases
- Broken contracts (function signatures changed without updating callers)
- Race conditions or concurrency issues
Security
- User input flowing to SQL, shell commands, file paths, or HTML without sanitization
- Hardcoded secrets, API keys, or credentials
- Overly broad permissions or missing auth checks
- Unsafe deserialization or eval usage
Performance
- O(n^2) or worse algorithms where O(n) is possible
- Unnecessary re-renders in React (missing memo/useMemo/useCallback)
- N+1 queries or unbounded data fetching
- Large allocations in hot paths
Architecture
- Does this change belong in this file/module?
- Are abstractions appropriate (not too much, not too little)?
- Does it follow existing patterns in the codebase?
- Will this be maintainable?
Style
- Naming clarity
- Dead code or TODOs left behind
- Missing types (in TypeScript)
- Inconsistent formatting (only flag if egregious)
Step 3: Write the Review
Output a structured review in this format:
## PR Review: <title> **Summary**: 1-3 sentence overview of what this PR does and whether it's ready. **Verdict**: APPROVE | REQUEST_CHANGES | COMMENT ### Findings #### <severity-emoji> <category>: <short description> **File**: `path/to/file.ts` L<line>-L<line> <explanation of the issue and why it matters> **Suggested fix**: \`\`\`diff - old code + new code \`\`\` --- (repeat for each finding)
Severity emojis:
- `[blocker]` - Must fix before merge. Bugs, security issues, data loss risks.
- `[warning]` - Should fix. Performance problems, poor patterns, missing edge cases.
- `[nit]` - Optional. Style, naming, minor improvements.
- `[praise]` - Good stuff. Call out well-written code (do this at least once).
Guidelines
- Be specific. Quote the code, give line numbers, explain why.
- Suggest fixes, not just problems. Include diff snippets.
- Don't nitpick formatting if there's a formatter configured.
- If the PR is large (>500 lines), focus on the riskiest files first.
- If the PR description explains a deliberate tradeoff, don't re-litigate it.
- Check the test coverage: are new code paths tested? Are edge cases covered?
Step 4: Post the Review (Optional)
Only post to GitHub if the user explicitly asks. Use:
# Post a review comment (not individual line comments) gh pr review <number> --repo <owner/repo> --comment --body "<review body>" # Or approve/request changes gh pr review <number> --repo <owner/repo> --approve --body "<review body>" gh pr review <number> --repo <owner/repo> --request-changes --body "<review body>"
For inline comments on specific lines:
gh api repos/<owner>/<repo>/pulls/<number>/comments \ -f body="<comment>" \ -f path="<file>" \ -f commit_id="$(gh pr view <number> --repo <owner/repo> --json headRefOid -q .headRefOid)" \ -F line=<line_number> \ -f side="RIGHT"
Always show the review to the user first and get confirmation before posting to GitHub.
Quick Mode
For small PRs (<100 lines changed), skip the full structure. Give a concise paragraph covering the key points and verdict.
Batch Review
When reviewing multiple PRs (e.g., "review all open PRs"):
gh pr list --repo <owner/repo> --json number,title,author,additions,deletions --jq '.[] | "#\(.number) \(.title) by @\(.author.login) (+\(.additions)/-\(.deletions))"'
Triage by size and risk, review the largest/riskiest first.
A 24/7 self-learning AI agent with an integrated IDE workspace that runs itself. Works with your existing Claude Code or OpenAI Codex subscription. No extra API key needed.
Repo: suitedaces/dorabot
Other skills on dorabot.
- /agent-swarm-orchestation
Master multi-agent orchestration using Claude Code's TeammateTool and Task system. Use when coordinating multiple agents, running parallel code reviews, creating pipeline workflows with dependencies, building self-organizing task queues, or any task benefiting from
Open skill - /github
Interact with GitHub using the `gh` CLI. Use `gh issue`, `gh pr`, `gh run`, and `gh api` for issues, PRs, CI runs, and advanced queries.
Open skill - /himalaya
CLI to manage emails via IMAP/SMTP. Use `himalaya` to list, read, write, reply, forward, search, and organize emails from the terminal. Supports multiple accounts and message composition with MML (MIME Meta Language).
Open skill - /macos
Control macOS via AppleScript/osascript. Manage windows (move, resize, tile), apps (launch, quit, focus), system (volume, dark mode, notifications), Spotify, browsers, Calendar, Reminders, Finder, and clipboard. Use when the user asks to control their Mac, arrange windows,
Open skill - /meme
Generate memes using the free memegen.link API. Create image memes from 200+ templates with custom top/bottom text, or use textual meme formats.
Open skill - /nano-banana
Generate and edit images using the Gemini API. Text-to-image, image editing, multi-turn iteration, 4K resolution, search grounding.
Open skill

