Skill Security — 安全审计扫描器,帮助你快速发现 Skill 中的安全风险。轻量 SAST 污点追踪(Python AST + JS 词法近似,source→sink 证据链降误报)、规则引擎(YAML 规则包热插拔扩展)、社区规则(schema 校验 + 来源记录 + 签名验证)、提示注入 ML 语义检测(ONNX + 正则降级)、系统级行为捕获(eBPF Linux / ETW Windows)、动态沙箱执行扫描、供应链风险分析、CVE 离线缓存、恶意 Skill 指纹库、健康度与合规检查(质量+结构+权限合并)、全局排除配置、CI/CD 集成、JSON/HTML/SARIF 报告生成。
Documents
secure-skill
Try itScan skills in a project directory for security issues and generate a markdown table report, then install skills from a local registry. Combines static analysis of code and markdown files with supply chain checks. Use when auditing a skills directory, generating a security summary table, or installi
What it does
Scan skills in a project directory for security issues and generate a markdown table report, then install skills from a local registry. Combines static analysis of code and markdown files with supply chain checks. Use when auditing a skills directory, generating a security summary table, or installing skills from a local registry.
The skill document
Secure Skill
Tier: POWERFUL Category: Engineering / Security
Overview
Scans all skills found under a target directory for security issues — covering code execution risks, prompt injection patterns, and supply chain vulnerabilities — and produces a structured markdown table report summarising the verdict for each skill. Also installs skills from a local registry into a project's .claude/skills/ directory.
Scripts
scripts/scan_skills.py— scans all skill subdirectories and writes a markdown reportscripts/install_skills.py— installs skills from a localregistry.jsoninto a project
Usage: Scan Skills
python scripts/scan_skills.py [--output report.md] [--strict] [--json]
Examples:
# Scan all skills in a directory and write report to scan_report.md
python scripts/scan_skills.py .claude/skills
# Scan with strict mode (HIGH findings count as FAIL) and save to custom path
python scripts/scan_skills.py .claude/skills --output security_report.md --strict
# Output raw JSON results
python scripts/scan_skills.py .claude/skills --json
Usage: Install Skills
Open scripts/install_skills.py and edit the DEFAULT_SKILLS list near the top:
DEFAULT_SKILLS = [
"skill-name-1",
"skill-name-2",
]
Then run:
python scripts/install_skills.py
To install specific skills without editing the file:
python scripts/install_skills.py skill-a skill-b skill-c
Report Format
The scan report is a markdown table with one row per skill:
| Skill | Code | Prompt | Supply Chain | Overall | Critical | High | Info |
|-------|------|--------|--------------|---------|----------|------|------|
| my-skill | PASS | PASS | PASS | PASS | 0 | 0 | 2 |
| bad-skill | FAIL | WARN | PASS | FAIL | 1 | 1 | 0 |
Overall verdict is the worst of the three individual verdicts. Exit code is 1 if any skill has a FAIL verdict, 0 otherwise.
Prerequisites
For scanning: The target directory should contain skill subdirectories, each with a SKILL.md file.
For installation: A registry.json file must exist in the skill root directory (same folder as this SKILL.md). Each entry needs "name" and "path" fields:
[
{ "name": "my-skill", "path": "my-skill" },
{ "name": "another-skill", "path": "another-skill" }
]
Behavior Notes
- Scanning reads files only; no files are modified or deleted during a scan
- All file access stays within the target directory tree
- No network connections are made at any point
- Installation stops with an error if a destination skill directory already exists
- All skills in a batch install are attempted; the final exit code reflects any failure
- The scanner skips fenced code blocks in markdown files to reduce false positives
Related skills
Audit installed AI agent skills against the OWASP Agentic Skills Top 10 and emit text, JSON, SARIF, or HTML reports.
Security scanner for AI agent skills. Detects hardcoded secrets, unsafe code execution, prompt injection, and malware patterns in under 50ms. Scan before you...
Run a security review workflow for Codex or ClawHub skills before activation or publication. Use when the user needs to inspect skill files, scripts, permiss...
Automated security audit for AI agent skills. Use BEFORE installing any skill from ClawHub, GitHub, or other sources. Scans SKILL.md + all files for 30+ red...
Use when asked to security-audit a repository, find vulnerabilities to fix, check for leaked secrets, review dependencies for known CVEs, or harden a project...