The Agent Skills Directory

[COMMAND_EXECUTION] (MEDIUM): The skill requests root scope and the Bash tool to execute scripts/lighthouse_audit.py. Granting root privileges for web performance profiling exceeds standard security practices and increases the impact of any script vulnerabilities.
[PROMPT_INJECTION] (MEDIUM): High-risk indirect injection surface through the processing of external URLs. 1. Ingestion point: The scripts/lighthouse_audit.py script takes a user-supplied or external URL as a parameter. 2. Boundary markers: No markers or instructions to ignore embedded instructions in the fetched data are present. 3. Capability: The agent has access to Bash and root scope while processing this external data. 4. Sanitization: There is no evidence of URL validation or output sanitization in the documentation.

performance-profiling