agentsclimarketplace

V8 hooks

Skill diguike/book-claude-skill/skills/code-review-snapshots/v8-hooks

Claude Code Skill 开发指南

Install
npx -y skills add diguike/book-claude-skill --skill v8-hooks

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

审查代码的质量、安全性和可维护性。当用户说'review 这段代码'、'帮我看看这个 PR'、'检查一下代码质量'、'这段代码有没有问题'时使用。

SKILL.md

2.9 KB, as published. Nobody here has run it

PR 上下文

!gh pr view $0 --json title,body,labels,changedFiles

变更概览

!gh pr diff $0 --stat

完整 diff

!gh pr diff $0

审查指令

基于以上 PR 信息,进行代码审查。

规则加载

始终加载 rules/base.md

根据代码内容,额外加载相关规则:

严重度标准

  • 🔴 Critical:必须修复才能合并。安全漏洞、数据丢失风险、线上必现 bug。
  • 🟡 Warning:建议修复。影响可维护性、潜在的边界 bug、性能隐患。
  • 🔵 Suggestion:可以考虑。更优雅的写法、微小的性能优化、代码风格。

不要把所有问题都标 Critical——只有真正阻塞合并的才算。

参考知识

根据技术栈,按需读取:

输出格式

严格按以下格式输出:

问题列表

#文件行号严重度问题描述修复建议

总结

  • 总体评分:X/10
  • 主要风险:一句话概括最大的问题
  • 亮点:值得肯定的做法
  • 建议优先级:先修什么,后修什么

自我评分

审查完成后,对本次审查质量进行评分(诚实评估,不要虚高):

  • coverage(0-1):你审查了变更文件的多大比例?
  • accuracy(0-1):你提出的问题中,有多少是真实存在的问题(而非过度警告)?
  • depth(0-1):你是否发现了需要理解业务逻辑才能看出的深层问题?

以 JSON 格式输出 selfScore。

审查后处理

审查完成后:

  1. 将审查结果保存到临时文件 /tmp/review-output.md
  2. 运行 npx tsx ${CLAUDE_SKILL_DIR}/scripts/collect-metrics.ts /tmp/review-output.md 生成统计
  3. 将统计结果和 selfScore 传给 npx tsx ${CLAUDE_SKILL_DIR}/scripts/append-log.ts data/review-metrics.jsonl '<统计JSON>' 记录日志

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.