菜单
Review a GitHub pull request thoroughly — analyze diff for correctness, security, breaking changes, code quality, and AI-slop patterns. Supports --fix to auto-remediate findings and --reply to post the review back to GitHub via the gh CLI as a formal review.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/hoanghd218/ZaloCRM --skill ak-review-pr命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
Search library/framework documentation via llms.txt (context7.com). Use for API docs, GitHub repository analysis, technical documentation lookup, latest library features.
Interview-driven advisory skill. Analyzes a prompt or URL (GitHub issue, spec, doc), scouts the codebase, interviews the user one question at a time to reframe the problem into exact requirements and goals, then delivers honest advice: what to do, what to avoid, better alternatives, benefits and trade-offs. Use when the user asks for advice, a second opinion, a sanity check, requirement reframing before committing to a direction, or wants an existing plan or design pressure-tested (grilled) one question at a time.
Browser and desktop automation through the agent-browser CLI. Use for long autonomous browsing, compact page snapshots, screenshots, form filling, login/test flows without real Chrome profile state, scraping, exploratory QA, cloud browsers, Electron apps, Slack automation, and bug hunts.
基于 SOC 职业分类
| name | ak:review-pr |
| description | Review a GitHub pull request thoroughly — analyze diff for correctness, security, breaking changes, code quality, and AI-slop patterns. Supports --fix to auto-remediate findings and --reply to post the review back to GitHub via the gh CLI as a formal review. |
| user-invocable | true |
| when_to_use | Invoke to review a GitHub PR by number/URL, optionally fix findings, optionally post the review back to GitHub. |
| category | utilities |
| keywords | ["pr","pull request","review","github","gh","fix","reply","anti-slop","ai-slop"] |
| argument-hint | <PR number or URL> [--fix] [--reply] |
| allowed-tools | ["Bash(gh pr view *)","Bash(gh pr diff *)","Bash(gh pr checks *)","Bash(gh pr review *)","Bash(gh pr comment *)","Bash(gh api *)","Bash(gh auth status *)","Bash(command *)","Bash(git log *)","Bash(git diff *)","Bash(git status *)","Bash(git branch *)","Bash(git rev-parse *)","Bash(git add *)","Bash(git commit *)","Bash(git push *)","Bash(date *)","Read","Edit","MultiEdit","Write","Glob","Grep","Task"] |
| metadata | {"author":"agentkit","version":"2.0.0"} |
Review PR $ARGUMENTS in this repository.
--fix): review, fix all actionable findings, commit+push, then re-review. Repeat until no actionable findings remain.--reply): after the review (or after the fix loop converges), post the review back to the PR via gh pr review.Flags compose: review-pr 123 --fix --reply runs the fix loop and posts the final re-review at the end. Flag order does not matter.
Derive PR_REF from $ARGUMENTS by stripping --fix and --reply flags:
!`PR_REF="$(printf '%s' "$ARGUMENTS" | sed -E 's/(^|[[:space:]])(--fix|--reply)([[:space:]]|$)/ /g; s/^[[:space:]]+//; s/[[:space:]]+$//')" && printf 'PR_REF=%s\n' "$PR_REF"`
Detect flags (the substring match below is intentional — flags may appear in any order):
--fix present → fix-loop mode active--reply present → reply mode activePR metadata:
!`PR_REF="$(printf '%s' "$ARGUMENTS" | sed -E 's/(^|[[:space:]])(--fix|--reply)([[:space:]]|$)/ /g; s/^[[:space:]]+//; s/[[:space:]]+$//')" && gh pr view "$PR_REF" --json title,body,author,baseRefName,headRefName,files,additions,deletions,changedFiles`
PR diff:
!`PR_REF="$(printf '%s' "$ARGUMENTS" | sed -E 's/(^|[[:space:]])(--fix|--reply)([[:space:]]|$)/ /g; s/^[[:space:]]+//; s/[[:space:]]+$//')" && gh pr diff "$PR_REF"`
CI check status:
!`PR_REF="$(printf '%s' "$ARGUMENTS" | sed -E 's/(^|[[:space:]])(--fix|--reply)([[:space:]]|$)/ /g; s/^[[:space:]]+//; s/[[:space:]]+$//')" && gh pr checks "$PR_REF" 2>/dev/null || echo "No checks found"`
Diff stat (use to gauge scope vs description claims):
!`PR_REF="$(printf '%s' "$ARGUMENTS" | sed -E 's/(^|[[:space:]])(--fix|--reply)([[:space:]]|$)/ /g; s/^[[:space:]]+//; s/[[:space:]]+$//')" && gh pr diff "$PR_REF" --name-only 2>/dev/null | head -50`
Perform a thorough code review of this PR. Follow these steps:
additions/deletions/changedFiles — a wide gap is itself a signal (see anti-slop reference)Correctness
Security
Breaking changes
Code quality (anti-slop — terse checklist) LLM-assisted PRs commonly introduce code that runs fine but pollutes the codebase. Scan the diff for these high-signal patterns:
utils/, helpers/, lib/common/, *manager.ts) without a clear domain anchorcatch (e) { console.log(e) } or catch { return null }// increment counter next to counter++)chunk, range, groupBy) when language or existing dep covers itany widening, @ts-ignore, // eslint-disable introduced to silence (not fix) warningsLoad the full taxonomy in references/anti-ai-slop.md when ANY of:
utils//helpers//lib/common/, ORThe reference covers: structural slop, micro slop, process slop, how to phrase the finding without becoming an AI-witch-hunt, when NOT to flag, and stack-specific appendix (Go, React/TS, Tailwind).
Project-specific compliance
CLAUDE.md, AGENTS.md, docs/code-standards.md, docs/system-architecture.md if presentreferences/project-rules-example.md for a worked example of project-specific compliance rules (Go gateway, React/Tailwind UI)Testing
Present your review as:
Summary: 1-2 sentence overview of what the PR does.
Risk level: Low / Medium / High — based on scope, complexity, and breakage potential.
Findings: List issues found, categorized by severity:
Anti-slop severity rule: structural slop (new dumping-ground file, parallel reimpl, abstraction with one caller, schema change without migration, large file growth) → Important. Micro slop (over-comments, defensive paranoia, one-line wrappers) → Suggestion. This keeps
--fixfrom churning the diff with cosmetic rewrites the original author won't recognize.
Verdict: One of:
--fix)If $ARGUMENTS contains --fix, follow this loop after the review steps above:
Activate ak:fix --auto with the full findings list and PR context:
ak:fix --auto "Fix all actionable findings from review-pr <PR_REF>: <finding summary>"
Pass the exact evidence:
ak:fix performs its own scout, diagnose, implementation, verification, and prevention flow. Do not bypass its hard gates.
After ak:fix verifies the fixes, activate:
ak:git cp
This stages, commits, and pushes the fixes to the PR head branch. Do not run ak:git cp if verification failed, secrets are detected, or the working tree contains unrelated user changes.
After the push succeeds, activate review-pr <PR_REF> --fix again (carrying --reply if it was originally set) and repeat the loop.
Stop only when one of:
ak:fix is blocked by a missing user/business decisionak:fix cannot resolve without user inputFinal output for --fix mode:
--reply)If $ARGUMENTS contains --reply, post the review back to GitHub as a formal review after the review (review-only) or after the fix loop converges (--fix).
Run these checks. On any failure, fall back to printing the review locally and warn the user — never fail the whole skill:
command -v gh >/dev/null 2>&1 || { echo "gh CLI not installed — printing review locally"; exit 0; }
gh auth status >/dev/null 2>&1 || { echo "gh not authenticated — printing review locally"; exit 0; }
Construct the full markdown body containing the summary, risk level, findings (by severity), and verdict. Append a single-line footer for traceability:
*Posted by the installed review-pr skill at <ISO-8601 UTC timestamp>*
Use date -u +"%Y-%m-%dT%H:%M:%SZ" for the timestamp.
Length cap: GitHub limits comment bodies to ~65,536 chars. If the body exceeds 60,000 chars, truncate the Findings section and append [truncated — N findings omitted; see local output] so the reviewer knows to consult the full chat output.
| Verdict | gh command |
|---|---|
| Approve | gh pr review "$PR_REF" --approve --body-file - |
| Request changes | gh pr review "$PR_REF" --request-changes --body-file - |
| Comment | gh pr review "$PR_REF" --comment --body-file - |
Pipe the body via stdin to avoid shell-quoting issues with backticks and code blocks.
GitHub blocks approving your own PR. If gh pr review --approve exits non-zero with a self-review error (HTTP 422, message matching "Can not approve your own pull request"), retry as a neutral formal review:
gh pr review "$PR_REF" --comment --body-file -
The review still lands in the timeline; the verdict text inside the body still reads "Approve". Note the downgrade in the chat output.
--fixIn --fix --reply mode, post only the final re-review when the loop converges. Iteration history lives in the commit log; the PR conversation stays clean.
If the loop terminates due to a blocker (non-converging, ak:fix blocked, CI unresolvable), still post the final review — but the verdict will reflect remaining findings (likely Request changes or Comment), and the body should include the blocker so the human reviewer knows where to take over.
V1 does not dedupe. Re-running review-pr 123 --reply posts a fresh review each time. The traceability footer (step 2) is the seed for future dedup work but is not consumed here.
After all modes complete, report to the chat:
--fix ran--fix ran--reply succeeded, fell back, or printed-locally