The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/security_scan.py uses subprocess.run to execute the npm audit command. This is a legitimate security analysis operation performed with a static argument list, which prevents command injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted third-party project files, creating an inherent surface for indirect prompt injection where malicious content in scanned files could attempt to misdirect the AI agent's analysis.
Ingestion points: Target project files located in the directory path provided to scripts/security_scan.py and accessed via the Read and Glob tools.
Boundary markers: The skill does not implement explicit prompt delimiters, but it uses a separate script to produce structured JSON findings, which provides separation between raw data and agent logic.
Capability inventory: The skill possesses file read access, pattern matching capabilities, and the ability to execute the local npm CLI tool.
Sanitization: The scanning script uses fixed regular expressions to extract specific security findings, reducing the risk of malicious data being interpreted as instructions.

vulnerability-scanner