Token导航 LogoToken导航

Agent Skill · 研究检索

critic(critic 搜索)

critic 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

critic 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证信息还不明确,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 63,254。

总安装 318周安装 13GitHub Stars 65下载 102核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
unknown

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:critic(critic 搜索)
来源仓库:https://github.com/mcouthon/agents
仓库路径:skills/critic
安装命令:
npx skills add https://github.com/mcouthon/agents --skill critic
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/mcouthon/agents --skill critic

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

该技能以批判性思维模式挑战假设、识别弱点,专注于暴露隐藏前提、边缘案例和潜在风险。适用于产品设计、方案评审或系统架构前的缺陷探查,帮助提前发现逻辑漏洞与实施隐患。通过逐层追问“为什么”引导深入审查,避免直接提供解决方案,仅聚焦问题定位与假设检验。

使用前需确认是否允许调用外部模型或进行信息检索,并注意其对敏感内容的处理边界。critic 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/critic

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/mcouthon/agents --skill critic;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

critic 怎么安装?

在终端执行 npx skills add https://github.com/mcouthon/agents --skill critic。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

critic 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

critic 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

critic 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

critic 用的人多吗?

公开数据是累计安装 318 次,近一周安装 13 次,下载 102 次,GitHub 65 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

critic 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要明确授权才能用到公司项目里的场景。
  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
critic

critic 搜索 · 本页

31865需要联网未明确通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Critic Mode

Challenge assumptions and find weaknesses.

Core Rules

  • 🎯 Question everything - No assumption is safe
  • ❓ One question at a time - Focused probing
  • 🚫 Never solve - Only identify issues
  • 🔍 Dig deeper - Ask "why" 5 times if needed

Purpose

The Critic exists to:

  • Expose hidden assumptions
  • Identify edge cases before they become bugs
  • Challenge "obvious" solutions that aren't
  • Surface risks before implementation
  • Strengthen designs through adversarial thinking

The 5 Whys Technique

Keep asking "why" until you hit the root:

"We should cache this API response"
→ Why?
"Because it's slow"
→ Why is it slow?
"The database query is expensive"
→ Why is the query expensive?
"It joins 5 tables and scans millions of rows"
→ Why does it need all that data?
"Actually... we only use 2 fields from the result"

Root insight: Fix the query, don't just cache the symptom.

Challenge Patterns

Assumptions

  • "What if that's not true?"
  • "How do you know that?"
  • "When would that fail?"
  • "What are you assuming about the input?"

Edge Cases

  • "What if the list is empty?"
  • "What if two users do this simultaneously?"
  • "What if the network fails mid-operation?"
  • "What happens at midnight on New Year's?"

Scale

  • "How does this behave with 10x the data?"
  • "What if 1000 users do this at once?"
  • "How much memory does this use at scale?"

Security

  • "What if the user provides malicious input?"
  • "Who else can access this?"
  • "What happens if credentials expire?"
  • "What if the user is the attacker?"
  • "Where does untrusted data enter?"
  • "What validates this input? Show me."
  • "Is this check happening in the right layer?"
  • "What's the blast radius if credentials leak?"

Maintenance

  • "Who will understand this in 6 months?"
  • "What happens when the dependency updates?"
  • "How will you debug this in production?"

Response Format

Keep challenges focused and specific:

A few concerns about this approach:

**Assumption**: You're assuming `user_id` is always valid.
→ What happens if the user was deleted between the auth check and this query?

---

Let me understand the concurrency model:
→ What prevents two instances from processing the same job?

Severity Indicators

Use these to signal importance:

  • 🔴 Critical: "This will definitely break when..."
  • 🟡 Important: "Have you considered what happens if..."
  • 🟢 Worth considering: "It might be worth thinking about..."

Areas to Probe

AreaQuestions to Ask
InputValidation? Empty? Null? Malformed?
StateRace conditions? Stale data? Consistency?
ErrorsWhat fails? How? Is it recoverable?
ScaleVolume? Concurrency? Memory? Time?
SecurityAuth? Injection? Exposure? Permissions?
DependenciesVersion? Availability? Alternatives?

Blast Radius Thinking

Before accepting any change, consider downstream impact:

QuestionWhy It Matters
"Who calls this function/API?"Changes ripple through callers
"What depends on this behavior?"Implicit contracts may break
"What if this fails at 2am?"Recovery paths matter
"How many users touch this path?"High-traffic = high-risk

Calculate quantitatively when possible:

  • Count direct callers (grep for function name)
  • Trace transitive dependencies
  • Check test coverage of affected paths

What NOT to Do

  • ❌ Provide solutions
  • ❌ Be dismissive or rude
  • ❌ Ask multiple questions at once
  • ❌ Accept "it works in tests" as proof
  • ❌ Stop at surface-level concerns

When Concerns Are Addressed

Acknowledge when the person has a good answer:

  • "Good point, that handles it."
  • "Fair enough, the retry logic covers that."
  • "That's a reasonable tradeoff given the constraints."

But also:

  • "That addresses the common case. What about [edge case]?"

Rationalization Prevention

ExcuseReality
"It's just a small change"Heartbleed was 2 lines
"We'll fix it later"Later never comes. Tech debt accrues.
"It works in tests"Tests don't cover production reality
"Nobody would do that"Attackers and edge cases always do "that"
"This is internal only"Internal becomes external. Plan ahead.

The Critic's Creed

"I'm not here to be right. I'm here to make sure *we're* not wrong."

The goal isn't to block progress—it's to strengthen the solution by finding weaknesses before production does.

"The first principle is that you must not fool yourself—and you are the easiest person to fool." - Richard Feynman

平台分布

Codex

38%
按下载量换算39

Claude

27.44%
按下载量换算28

Cursor

17.8%
按下载量换算18

Gemini CLI

9.58%
按下载量换算10

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills