Token导航 LogoToken导航TokenDH.com
研究检索操作浏览器github未标认证来源可访问许可证需确认审计提醒

audit审核

Agent Skill

用于辅助安全审计、权限检查、凭据风险、认证流程和常见漏洞排查。它适合让 Agent 梳理敏感配置、检查依赖风险、分析鉴权逻辑或生成安全复核清单。使用时不能把工具输出直接当最终结论,涉及密钥、令牌、用户数据或生产系统时,应先确认最小权限、脱敏方式和操作边界。

总安装

264

周安装

11

GitHub Stars

公开资料未说明

下载量

88
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

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:audit(审核)
来源仓库:https://github.com/cerico/macfair
仓库路径:skills/audit
安装命令:
npx skills add https://github.com/cerico/macfair --skill audit
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/cerico/macfair --skill audit

简介

Security Audit 智能分析 npm audit 结果,区分真实威胁与传递依赖噪音。

  • 适用于部署前安全检查、依赖更新后的漏洞排查及定期安全审查。
  • 能识别直接依赖与传递依赖的风险,并提供具体修复建议。
  • 安装前需确认是否会执行命令或读取 package.json,涉及生产环境时应先脱敏处理。
  • 输出仅为辅助结论,不能直接作为最终安全判断依据。

SKILL.md

Security Audit

Intelligently analyze npm audit results. Distinguishes real threats from transitive dependency noise and recommends specific actions.

When to Use

  • After make audit shows vulnerabilities
  • Before deploying to production
  • When updating dependencies
  • Periodic security check (integrated into /preflight)

Instructions

1. Run Audit

pnpm audit --json 2>/dev/null || true

If no package.json exists, report "No Node.js project found" and exit.

2. Parse Each Vulnerability

For each vulnerability, determine:

A. Direct vs Transitive

Look at the via and effects fields:

  • Direct: Your package.json depends on the vulnerable package
  • Transitive: A dependency of a dependency is vulnerable
# Check if package is a direct dependency
grep -q '"<package-name>"' package.json && echo "DIRECT" || echo "TRANSITIVE"

B. Production vs Dev

# Check if it's a dev dependency
grep -A5 '"devDependencies"' package.json | grep -q '"<package-name>"' && echo "DEV" || echo "PROD"

For transitive deps, trace up to find the root:

pnpm why <vulnerable-package>

C. Exploitability

Consider:

  • Is the vulnerable code path actually used?
  • JWT vulnerabilities don't matter if you use Clerk
  • XSS in markdown parser doesn't matter if you sanitize output
  • Server-only packages aren't exposed to client attacks

3. Check for Existing Overrides

Before recommending fixes, check if overrides are already in place:

# Check package.json for pnpm overrides
grep -A20 '"pnpm"' package.json | grep -A15 '"overrides"'
Note: This checks pnpm-style overrides only. npm uses "overrides" at root level, yarn uses "resolutions".

If an override exists for the vulnerable package:

  • Check if the override version is >= patched version
  • If yes: vulnerability should be resolved (verify with pnpm list <package>)
  • If no: override exists but needs version bump

4. Check for Fixes

For each vulnerability:

# Check if parent package has update
pnpm outdated <parent-package> 2>/dev/null

# Check latest version
pnpm info <parent-package> version

5. Classify

CategoryCriteriaAction
CriticalDirect + Production + ExploitableFix immediately
HighDirect + ProductionUpdate when convenient
MediumTransitive + ProductionMonitor, check for parent updates
LowDev dependency onlyCan ignore, update eventually
OverriddenHas pnpm override >= patched versionVerify resolved, no action needed
NoiseTransitive + Dev + Not exploitableIgnore

6. Recommend Actions

For each non-noise vulnerability:

If direct dependency:

pnpm update <package>@latest

If transitive (parent has fix):

pnpm update <parent-package>@latest

If transitive (no fix available):

  • Check GitHub issues on parent package for timeline: gh search issues --repo <owner/repo> "<vulnerable-package>" --state open
  • Consider override (last resort):
// package.json
"pnpm": {
  "overrides": {
    "<vulnerable-package>": ">=<fixed-version>"
  }
}

If not exploitable in your context:

  • Document why it's not a risk
  • Revisit on next audit

Output Format

## Security Audit Report

**Summary:** X vulnerabilities found, Y actionable

### Critical (fix now)
- **<package>** (<severity>) - <vulnerability>
  - Type: Direct / Production
  - Fix: `pnpm update <package>@<version>`

### Should Fix
- **<package>** (<severity>) - <vulnerability>
  - Type: Transitive via `<parent>`
  - Fix: `pnpm update <parent>@<version>`

### Monitor
- **<package>** (<severity>) - <vulnerability>
  - Type: Transitive via `<parent>`
  - Status: No fix available, tracked in <github-issue-link>

### Already Resolved (via override)
- **<package>** (<severity>) - <vulnerability>
  - Override: `"<package>": ">= <version>"` in package.json
  - Verified: `pnpm list <package>` shows <resolved-version>

### Noise (can ignore)
- **<package>** (<severity>) - <vulnerability>
  - Reason: Dev dependency only / Not exploitable because <reason>

### Recommended Actions
1. {specific command to run}
2. {specific command to run}

### Next Audit
Run `/audit` again after updates to verify fixes.

Common Scenarios

"Vulnerability in Prisma's dev tools"

Prisma bundles dev tools (Studio) with their own dependencies. These don't affect your production app.

  • Verdict: Noise (dev tooling, not in your code path)

"Vulnerability in testing framework"

Jest, Vitest, Playwright vulnerabilities are dev-only.

  • Verdict: Low priority (update when convenient, doesn't affect prod)

"JWT vulnerability but using Clerk"

If you don't use the vulnerable JWT code path, it's not exploitable.

  • Verdict: Noise (not exploitable in your context)

"XSS in markdown parser"

Only matters if you render untrusted markdown without sanitization.

  • Verdict: Check if you sanitize. If yes, noise. If no, critical.

Notes

  • pnpm audit returns exit code 1 when vulnerabilities exist (expected)
  • Override caveat: pnpm audit may still report vulnerabilities even with overrides, because it checks declared versions in lockfile metadata. Verify actual resolution with pnpm list <package>
  • Some packages will never be fixed (abandoned) - consider alternatives
  • Overrides can break things - use sparingly and test thoroughly
  • After adding/changing overrides, run pnpm install to apply

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35.32%
按下载量换算31

Claude

28.97%
按下载量换算25

Cursor

19.04%
按下载量换算17

Gemini CLI

10.17%
按下载量换算9

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

操作浏览器

该 Skill 可能涉及浏览器控制能力,使用时可能读取或操作网页内容,需要在受控环境中确认权限边界。

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。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills