How this skill is triggered — by the user, by Claude, or both
Slash command
/pr:review [PR number or URL][PR number or URL]This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Review a GitHub pull request using the `gh` CLI. Examine it as a thorough human reviewer would.
Review a GitHub pull request using the gh CLI. Examine it as a thorough human reviewer would.
Identify the PR:
gh pr view --json number,urlFetch PR details by running:
gh pr view <number> --json title,body,baseRefName,headRefName,files,additions,deletions,commits,reviews,comments,url
Read the PR description carefully:
curl via bash) to understand the context.Examine the diff:
gh pr diff <number>
If the diff is very large, also look at the file-level summary:
gh pr diff <number> --stat
and then examine the most important files in detail.
Conduct the review by evaluating:
Present the review to the user:
Offer next steps. Ask the user which of these to do:
Option A: Post review comments on the PR
gh pr review <number> to submit the review.[AI Review] so it is obvious the feedback was AI-generated.--comment for suggestion-level reviews, --request-changes if there are blocking issues.Option B: Check out the branch and implement fixes locally
gh pr checkout <number>Option C: Do nothing (just keep the review as conversation output).
gh, always label them as AI-generated.Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub wincent/wincent-agent-plugins --plugin pr