Scan staged ClawHub skills for prompt injections, secrets, and malware before install.
npx skills add https://github.com/LudwigAJ/swell --skill skill-guard-ludwigaj
Please help me install this Agent Skill. Skill: skill-guard Source: https://github.com/LudwigAJ/swell/tree/main/.factory/skills/skill-guard Command: npx skills add https://github.com/LudwigAJ/swell --skill skill-guard-ludwigaj
Skill-guard helps you avoid installing malicious or unsafe ClawHub skills by checking the actual skill contents before they reach your workspace.
Ask the assistant to scan the ClawHub skill named some-skill and only install it if the pre-flight security check passes.
High-intent search queries and answers about installing and using this skill.
To scan ClawHub skills for prompt injection, skill-guard stages the skill in a temporary directory and runs mcp-scan to detect prompt injections, secret leaks, and malware patterns before the files reach your workspace.
You can check new AI skills for malware and secret leaks by staging them in a temporary directory using skill-guard, which keeps your real skills folder untouched while it analyzes the files for security threats.
Yes, you need uvx and mcp-scan to run a pre-install security check, as skill-guard requires both dependencies to fetch skills into a temporary staging directory and analyze the files for AI-specific threats.
When a security scan detects threats, suspicious skills are quarantined by skill-guard, providing you with a clear path to review the staged files, discard them, or install them anyway if you intentionally accept the risk.
After a pre-flight security check fails, skill-guard quarantines the unsafe skill in a staging directory, allowing you to review the detected prompt injections or malware patterns and discard the files before they reach your workspace.
Yes, you can install a quarantined skill if you intentionally accept the security risks, as skill-guard provides a clear path to proceed with the installation even after mcp-scan detects prompt injections or malware patterns.
mvillmowDetect security vulnerabilities and hardcoded secrets in code.
rbah31Arbitrate wiki proposals and apply merge, rewrite, discard, or defer decisions.
sigitpoerwoSearch, install, update, and publish agent skills via the ClawHub CLI.
oioio-spaceScan staged Git diffs for credentials, PII, and sensitive data.
