The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Downloads skill packages from the vendor-associated domain auth.clawdhub.com for security inspection and analysis.\n- [COMMAND_EXECUTION]: Executes a local Python-based security scanner and standard utilities like curl and unzip to process and inspect external code packages.\n- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it ingests and analyzes untrusted third-party code. \n
Ingestion points: The scan.py script reads the raw text content of files within the directory being vetted. \n
Boundary markers: Scanned content is reported via file and line references but is not isolated with specialized security delimiters in the output. \n
Capability inventory: The skill has command-line capabilities including curl, unzip, and grep for file management and searching. \n
Sanitization: Content extracted from the scanned files is not sanitized or filtered before being displayed to the agent.

skill-vetting