Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
v2.0.3 introduces .vettrignore file support, customizable vetting roots, and safer filesystem defaults. - Added support for .vettrignore files to exclude files/directories from scans using glob patterns. - Introduced allowCwd config option: by default, only os.tmpdir(), ~/.openclaw, and ~/Downloads are accepted as vetting roots; set allowCwd: true to scan under the current working directory. - Added startup verification for required .wasm artifacts with descriptive error messages when files are missing after install. - Added formal install spec to registry metadata declaring npm install command, required .wasm artifacts, and lifecycle script safety notes. - Updated documentation to explain filesystem scope, configuration, and .vettrignore format and usage.
What This Skill Does
Static analysis scanner that checks third-party OpenClaw skills for security risks before installation. It parses JavaScript and TypeScript source using tree-sitter AST analysis and regex to detect eval usage, shell injection, malicious dependencies, typosquatting, and prompt injection patterns. Supports local directories, remote URLs, and ClawHub registry slugs.
Tree-sitter AST parsing catches obfuscated patterns that regex-only scanners miss, including dynamic require calls, prototype pollution, and restructured eval equivalents across minified source.
When to use it
- Vetting a ClawHub skill before installing it on a production agent
- Scanning a skill shared as a URL for shell injection or eval patterns
- Checking a locally developed skill for accidental credential path exposure
- Detecting typosquatted dependency names in a third-party skill's package.json
- Auditing a skill's SKILL.md for dangerous permission declarations before granting access
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: vet the 'data-exfil' skill from ClawHub before installing
- 1Runs /skill:vet-clawhub --skill data-exfil to fetch the skill via the clawhub CLI
- 2Extracts the archive into a temp directory within allowed vetting roots
- 3Parses JS/TS files with tree-sitter to flag eval, spawn, and dynamic require nodes
- 4Scans package.json for known malicious packages, git/http dependencies, and lifecycle scripts
- 5Checks the skill name against known targets using Levenshtein distance for typosquatting
Report listing detected risks by category with file paths and line references for each finding
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.