Gh Review
Draft and post direct, unhedged GitHub PR reviews and issue comments, every item is a...
Install
npx skills add mastra-ai/mastra --skill gh-review -g -a claude-code -y- Repository
- mastra-ai/mastra29.3k
- Category
- Git and review
- Updated
- 3d ago
- License
- See repository
- Context cost
- Mediumabout 2,356 tokens
- Runs scripts
- No, instructions only
What it does
Draft and post direct, unhedged GitHub PR reviews and issue comments, every item is a required change with a pointer and reason, PR reviews land as "request changes" or "approve" (never a bare comment), and nothing is marked optional or deferred to a follow-up. Use whenever understand-pr or understand-issue (or any review) is about to draft or post to GitHub.
GitHub review voice
Load this the moment a review or diagnosis is about to become a GitHub comment. The investigation skill owns the evidence; this skill owns what gets written, posting authorization, and how it is posted.
Who is posting
Before drafting, run gh auth status once and record the active account for the target GitHub host in the review record. Do not dump environment variables or expose tokens.
Do not assume the answer. It may be the user's account or a bot account with its own identity; which one it is changes how you act:
- Self-review means the posting account is the PR author. Compare the login above to the PR's
author.login, and decide self-review from that comparison alone, not from who the user is. - Write in the posting account's voice. If it is not the author's account, do not refer to the PR as "mine".
- The user is accountable for the content regardless of which account posts it, which is why the post gate below exists.
Check once per session, not before every post.
The rule
Ask for the changes you want, and every one of them is required. Small requests are still requests, post them. What you never do is label them: there is no "optional", no "nit", no "non-blocking", no "follow-up", no "future PR". The author satisfies the list; the review is not a menu.
Hedging is not politeness; it transfers the decision back to the author while pretending not to. A reviewer who has evidence states the request.
Voice
Be direct about required changes and precise about what the evidence establishes:

