pr-code-review

📁 arctuition/skills 📅 Jan 29, 2026
4
总安装量
4
周安装量
#50656
全站排名
安装命令
npx skills add https://github.com/arctuition/skills --skill pr-code-review

Agent 安装分布

codex 4
opencode 3
gemini-cli 3
antigravity 2
claude-code 2
junie 2

Skill 文档

PR Code Review

Use this skill to review GitHub PRs end-to-end with the gh CLI, leave inline comments tied to specific code lines, and conclude with a severity summary.

Quick workflow

  1. Identify the PR context.
  2. Collect and read diffs.
  3. Analyze and draft findings.
  4. Post inline comments on the relevant code lines.
  5. Provide a severity summary (high/medium/low).

Step-by-step

1) Identify PR and scope

  • Resolve the PR number or URL and repository.
  • Capture metadata to anchor comments and fetch commit IDs.

Suggested commands (see references/gh-cli.md):

  • gh pr view <pr> --json number,title,headRefOid,baseRefName,headRefName,author,changedFiles,additions,deletions
  • gh pr view <pr> --json files --jq '.files[] | {path,additions,deletions}'

2) Read the diff efficiently

  • Start with file list and high-churn areas.
  • Use patch output when you need line numbers.
  • For larger PRs, break the diff into files and review sequentially.

Suggested commands:

  • gh pr diff <pr> --name-only
  • gh pr diff <pr> --patch
  • gh pr diff <pr> --patch --color=never > /tmp/pr.diff

3) Analyze and record findings

  • Classify each issue as High / Medium / Low severity.
  • For each issue, capture: file path, line number, summary, reasoning, and recommended fix.
  • Keep comments crisp and actionable.

4) Post inline comments in the PR

  • Prefer inline comments tied to exact file/line.
  • If inline comments are not possible, fall back to a general review comment.

Primary path (inline comment via GitHub API using gh api):

  • Use headRefOid as commit_id.
  • Use file path and line number from the patch output.

Fallback path (general review comment):

  • Use gh pr review --comment -b "..." or gh pr comment -b "...".

See references/gh-cli.md for examples.

5) Final summary by severity

Provide a final summary grouped by severity:

  • High: issues that can cause correctness, security, data loss, or outage.
  • Medium: issues that affect reliability, maintainability, or performance.
  • Low: minor style issues, low-risk refactors, or optional improvements.

Include counts and short titles of each issue, and mention any files that were not reviewed (if any).

Output format

  • Use short, direct sentences.
  • Always include file path and line number with each inline comment.
  • End with a severity summary in three sections: High / Medium / Low.

References

  • references/gh-cli.md for command patterns and inline comment API usage.