blindspot-finder

仓库创建 2026年7月26日最近提交 1 天前SkillHot 收录 1 天前

▸ 精选理由

引入对立人格视角，能发现常规审查易漏掉的问题点。

这个 Skill 做什么

通过多重敌对视角发现代码审查中的盲点与遗漏。

用四种“敌对”视角做对抗性代码审查：分别从正确性、可维护性、安全性和生产可操作性去挑毛病，要求每个视角都提出实质问题再下结论。合并 PR 前、评审可能走过场或代码审查后感觉疲劳想要二次把关时用。相比普通复审，它系统化模拟不同攻击/运维角色，能发现原作者容易忽略的盲点，比如日志、报警或回滚风险。

▸ 展开 SKILL.md 英文原文

Adversarial code review that finds the self-review blind spot by forcing four hostile perspective shifts before rendering a verdict. Use when reviewing a diff or PR before merge, when a review feels like it's about to be a rubber stamp, when Claude just said "looks good" and a second, harsher opinion is wanted, or after a long session where fatigue may be hiding bugs. Covers correctness/robustness, maintainability, security, and production operability (logging, alerting, rollback safety) — the last of which most adversarial reviewers skip.

开发编程代码审查盲点发现对抗性评估通用

Stars

Forks

仓库内 Skill

7 日增星

安装 / 使用

给你的 Agent 一句话（通用）

帮我安装这个 skill：https://raw.githubusercontent.com/tamasbege/staff-engineer-skills/main/plugins/staff-engineer-skills/skills/blindspot-finder/SKILL.md

或 curl 直取 SKILL.md

curl -fsSL "https://raw.githubusercontent.com/tamasbege/staff-engineer-skills/main/plugins/staff-engineer-skills/skills/blindspot-finder/SKILL.md"

SKILL.MD 节选查看完整文件 ↗

# Blindspot Finder

When you review code you just wrote (or just read), your judgment shares the same mental model that produced it — you notice what you expected to see, not what's actually there. This skill finds what that blind spot hides by forcing four hostile personas, each with a different fear and a different definition of "broken," and requiring every one of them to find something. There is no "LGTM" exit.

**Credit:** the three-persona structure (Saboteur, New Hire, Security Auditor) and the severity-promotion mechanic are adapted from the `adversarial-reviewer` skill in Alireza Rezvani's [claude-skills](https://github.com/alirezarezvani/claude-skills) `engineering-skills` plugin (

via SKILL·HOT · 数据来自 GitHub 公开信息 · 原文版权归作者所有