skill-audit
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [DATA_EXFILTRATION]: The skill accesses sensitive local configuration files such as
~/.claude.json,~/.claude/settings.json, and.mcp.jsonto perform its audit. It limits potential data exposure by only reporting high-level metadata (names, transport types, and sizes) and explicitly avoids printing sensitive contents like environment variables, credentials, or file bodies. \n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata (names and descriptions) from third-party skills and plugins. It implements a defensiveclean()function that uses a strict whitelist regex ([A-Za-z0-9._:@/-]) to filter these strings, ensuring that malicious instructions hidden in skill names cannot influence the agent's behavior.
Audit Metadata