The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a script (scripts/security_scan.py) that executes a system command to audit package dependencies.
Evidence: The script calls subprocess.run(["npm", "audit", "--json"], ...) to identify vulnerable Node.js packages.
Context: This is a legitimate use of command execution for security analysis purposes and is implemented safely without shell interpolation.
[INDIRECT_PROMPT_INJECTION]: The scanner processes external code and configuration files, creating a potential surface for indirect prompt injection if the files contain malicious instructions.
Ingestion points: scripts/security_scan.py reads and processes the content of all files within a target project directory provided by the user.
Boundary markers: Content is processed directly via regex without being wrapped in specific security delimiters or markers to warn the model about untrusted data.
Capability inventory: The skill has file system read access and the ability to execute the npm binary.
Sanitization: No sanitization is performed on the file contents beyond the regex matching patterns defined in the script.
Context: This ingestion of untrusted data is the primary purpose of the security scanner and does not indicate malicious intent.

vulnerability-scanner