跳到主内容
智客 ZICQ

技能库 智客分类:安全测试 skills-security-audit

Skills Security Audit

在安装前或定期对安全风险进行审计AI代理技能。 工作在克洛德代码,OpenClaw,以及所有平台上. 检测出即时注射,数据过滤,恶意命令,模糊代码,特权滥用,供应链风险,内存中毒,信任剥削等行为操纵. 在从任何市场安装第三方技能之前使用.

359 安装量

官方网址:skills.sh

技能介绍

先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。

做什么

在安装前或定期对安全风险进行审计AI代理技能。 工作在克洛德代码,OpenClaw,以及所有平台上. 检测出即时注射,数据过滤,恶意命令,模糊代码,特权滥用,供应链风险,内存中毒,信任剥削等行为操纵. 在从任何市场安装第三方技能之前使用.

何时用

- 在从市场安装任何第三方技能或插件之前

代理如何加载

按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:Skill Security Audit、Overview、When to Use、Security Check Categories、Audit Workflow、Phase 1: Determine Scan Scope。 其中含规范建议的小节:分步指令。

文件分析

文件分析:除 SKILL.md 外,正文引用了 references/security-rules.md,属于带资源的技能包,这些文件按需再读。

官方 description(原文)

Audit AI agent skills for security risks before installation or periodically. Works on Claude Code, OpenClaw, and all platforms. Detect prompt injection, data exfiltration, malicious commands, obfuscated code, privilege abuse, supply chain risks, memory poisoning, trust exploitation, and behavioral manipulation. Use before installing third-party skills from any marketplace.

Skill Security AuditOverviewWhen to UseSecurity Check CategoriesAudit WorkflowPhase 1: Determine Scan ScopePhase 2: Analyze Each FilePhase 3: Generate ReportReport FormatSingle Skill ReportSkill Security Audit ReportTarget: [skill-name] [version if available]

来源分类:skills.sh agent-skill

SKILL.md 与 Agent 调用

官方规范 ↗
name
skills-security-audit
description
Audit AI agent skills for security risks before installation or periodically. Works on Claude Code, OpenClaw, and all platforms. Detect prompt injection, data exfiltration, malicious commands, obfuscated code, privilege abuse, supply chain risks, memory poisoning, trust exploitation, and behavioral manipulation. Use before installing third-party skills from any marketplace.
  1. 发现技能客户端向 Agent 提供名称与描述目录。
  2. 匹配与调用用户指定或任务匹配后,载入 SKILL.md 指令。
  3. 按需加载按步骤读取参考文档、使用脚本与素材。
指令中引用的文件 · 1
  • references/security-rules.md

以下路径提取自原文;文件是否齐全请以来源仓库中的完整目录为准。

具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗

安装这个技能

Skills CLI ↗

先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。

该技能引用了附属文件,请从来源获取完整目录;仅复制 SKILL.md 可能缺少依赖。

交给 Agent 安装

复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。

把 Agent Skill「skills-security-audit」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-f274c63dc1705b8e-%E6%8A%80%E8%83%BD%E5%AE%89%E5%85%A8%E5%AE%A1%E8%AE%A1.html
请存为 .cursor/skills/skills-security-audit/SKILL.md 或 .claude/skills/skills-security-audit/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。
该技能还带 scripts/、references/、assets/ 等文件,请从 https://github.com/agentnode-dev/skills-security-audit 取完整目录,不要只建一个 SKILL.md。

GitHub 完整包 ↗

终端安装 · Skills CLI

需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。

npx skills add 'https://github.com/agentnode-dev/skills-security-audit' --list

npx skills add 'https://github.com/agentnode-dev/skills-security-audit' --skill 'skills-security-audit'

CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。

阅读排版

name: skills-security-audit description: > Audit AI agent skills for security risks before installation or periodically. Works on Claude Code, OpenClaw, and all platforms. Detect prompt injection, data exfiltration, malicious commands, obfuscated code, privilege abuse, supply chain risks, memory poisoning, trust exploitation, and behavioral manipulation. Use before installing third-party skills from any marketplace.

Skill Security Audit

Overview

Scan and audit AI agent skills, plugins, and tool definitions for security vulnerabilities across nine risk categories aligned with the OWASP Agentic AI Top 10 (ASI01 through ASI10). This skill works cross-platform with Claude Code, OpenClaw, and any AI agent platform that uses file-based skill definitions. Rather than relying on brittle regex patterns, it performs AI-powered semantic analysis to detect prompt injection, data exfiltration, obfuscated code, privilege escalation, supply chain attacks, memory poisoning, trust boundary violations, and behavioral manipulation. Each audit produces a structured risk report with severity ratings, evidence citations, and actionable remediation guidance.

When to Use

  • Before installing any third-party skill or plugin from a marketplace
  • When reviewing skills downloaded from OpenClaw, ClawHub, or other registries
  • Periodic audit of all installed skills and plugins
  • When a skill requests unusual permissions or behaves unexpectedly

Security Check Categories

| ID | Category | Severity | OWASP ASI | |----|----------|----------|-----------| | PI | Prompt Injection | CRITICAL | ASI01 | | DE | Data Exfiltration | CRITICAL | ASI02 | | CE | Malicious Command Execution | CRITICAL | ASI02, ASI05 | | OB | Obfuscated/Hidden Code | WARNING | — | | PA | Privilege Over-Request | WARNING | ASI03 | | SC | Supply Chain Risks | WARNING | ASI04 | | MP | Memory/Context Poisoning | WARNING | ASI06 | | TE | Human Trust Exploitation | WARNING | ASI09 | | BM | Behavioral Manipulation | INFO | ASI10 |

Load references/security-rules.md (relative to this file's directory) for detailed detection patterns, examples, and false positive guidance.

Audit Workflow

Phase 1: Determine Scan Scope

  • If user specifies a directory path, scan all files in that directory recursively.
  • If user says "scan installed", scan platform-specific skill directories:
    • Claude Code: ~/.claude/plugins/cache/
    • Cursor: ~/.cursor/extensions/ and project .cursorrules
    • Windsurf: ~/.codeium/windsurf/
    • Other platforms: ask user for the directory path.
  • If user provides a GitHub URL, use WebFetch to retrieve the repository content, or clone it locally.
  • Scan these file types: .md, .json, .js, .py, .sh, .ts, .yaml, .yml
  • List all files found and confirm with user before proceeding.

Important: Do NOT dispatch this audit to a subagent (Task tool). Subagents run in a sandboxed environment that cannot read ~/.claude/plugins/cache/ or other system directories. Always run the audit in the main conversation context.

Phase 2: Analyze Each File

  • Load references/security-rules.md (relative to this file's directory) for detailed detection patterns.
  • Read each file using the Read tool.
  • Check file content against all 9 categories (PI, DE, CE, OB, PA, SC, MP, TE, BM).
  • For each finding, record: rule ID (e.g., PI-001), severity, file path and line number, description, and recommended action.
  • Apply context-aware judgment — not every pattern match is a true positive.
  • When a single code block triggers multiple rules, report each applicable rule separately. Cross-category overlap (e.g., OB + CE + PI on the same line) increases confidence that the finding is a true positive.
  • Consider the skill's stated purpose when evaluating findings. A security auditing skill will naturally reference dangerous patterns.

Phase 3: Generate Report

  • Calculate risk score using the scoring formula below.
  • Output the structured report using the template below.
  • For batch scans of multiple skills, output a summary table at the end.

Report Format

Single Skill Report

When auditing one skill, output this full report:

## Skill Security Audit Report

### Target: [skill-name] [version if available]
### Risk Score: X.X/10 ([LEVEL])

---

### CRITICAL

- [PI-001] file.md:42 — Description of finding
  Risk: Why this is dangerous
  Action: Recommended response

### WARNING

- [OB-003] script.js:15 — Description of finding
  Risk: Why this is concerning
  Action: Recommended response

### INFO

- [BM-002] SKILL.md:88 — Description of finding
  Risk: Why this is worth noting
  Action: Recommended response

---

### Summary
- CRITICAL: N | WARNING: N | INFO: N
- Risk Score: X.X/10 — [Overall recommendation]

Batch Scan Report

When scanning multiple skills, use this compact format. Start with the summary dashboard, then show only skills with findings:

## Skill Security Audit — Batch Report

### Dashboard

| # | Skill | Score | Level | C | W | I | Top Finding |
|---|-------|-------|-------|---|---|---|-------------|
| 1 | skill-a | 0.0 | ✅ SAFE | 0 | 0 | 0 | — |
| 2 | skill-b | 2.4 | ⚠️ RISKY | 0 | 3 | 0 | [PA-001] Blanket permission grant |
| 3 | skill-c | 6.0 | 🔴 DANGEROUS | 2 | 1 | 1 | [DE-001] Reads ~/.ssh/id_rsa |
| 4 | skill-d | 8.2 | 🟣 MALICIOUS | 3 | 2 | 0 | [CE-003] curl | sh execution |

**Scanned: 4 skills | Clean: 1 | Needs review: 3**

---

### #3 skill-c — 6.0/10 🔴 DANGEROUS

| Rule | File:Line | Finding | Action |
|------|-----------|---------|--------|
| [DE-001] CRITICAL | lib/init.sh:14 | Reads `~/.ssh/id_rsa` | Remove sensitive file access |
| [DE-004] CRITICAL | lib/init.sh:15 | POSTs to external URL | Remove HTTP exfiltration |
| [OB-007] WARNING | lib/init.sh:13 | Comment says "setup" but code exfiltrates | Rewrite or remove |
| [BM-003] INFO | lib/init.sh:16 | Suppresses stderr output | Review necessity |

### #4 skill-d — 8.2/10 🟣 MALICIOUS
...

Batch format rules:

  • Dashboard table always comes first — gives the user an instant overview.
  • Only expand details for skills scoring above 0.0 (skip SAFE skills).
  • Use a compact table per skill instead of nested bullet lists.
  • Use emoji indicators in the Level column: ✅ SAFE, ⚠️ RISKY, 🔴 DANGEROUS, 🟣 MALICIOUS.
  • Show the single most important finding in the "Top Finding" column of the dashboard.

Scoring

Calculate risk score:

  • Each CRITICAL finding: +2.0 points
  • Each WARNING finding: +0.8 points
  • Each INFO finding: +0.2 points
  • Maximum score: 10.0

Risk levels:

  • 0.0–2.0: SAFE — No significant risks found.
  • 2.1–5.0: RISKY — Manual review recommended before use.
  • 5.1–8.0: DANGEROUS — Do not install.
  • 8.1–10.0: MALICIOUS — Confirmed malicious intent. Report to marketplace.

False Positive Guidance

  • Consider the skill's legitimate purpose before flagging.
  • A security auditing skill will naturally reference dangerous patterns — this is not malicious.
  • Development tools may legitimately need Bash access.
  • Look for intent, not just pattern presence.
  • When uncertain, report the finding with a note explaining the ambiguity.

相关技能

安全测试

Skill Vetter

人工智能特工的安全第一技能审查。 在安装来自 ClawdHub, GitHub 或其他来源的任何技能之前使用 。 检查红旗,许可范围,可疑图案.

安全测试

Moltguard

MoltGuard——由OpenGuardrails担任OpenClaw保安. 安装 MoltGuard 以保护你和你的人类免受迅速注射、数据过滤和恶意命令的影响。 资料来源:https://github.com/openguardrails/openguardrails/tre…

安全测试

Security Auditor

用于审查安全漏洞代码,实施认证流,审计OWASP Top 10,配置CORS/CSP头,处理秘密,输入验证,SQL注射预防,XSS保护,或任何与安全相关的代码审查.

安全测试

Skill Vetter

安装任何AI代理技能前的安全审查协议. 红旗侦取入证盗取,疏通密码,放出. 风险分类:低/中/高/高。 编制结构化审查报告。 不先运行此功能, 绝不安装不信任的技能 .