The Agent Skills Directory

[COMMAND_EXECUTION]: Executes standard shell commands and command-line tools to perform security audits on project files.
Evidence: Utilizes git grep, grep, npm audit, bun pm audit, and govulncheck to identify potential security issues in the codebase (SKILL.md).
[EXTERNAL_DOWNLOADS]: Performs dependency audits that involve connecting to official package registries to retrieve vulnerability information.
Evidence: Suggests using uv run pip-audit, npm audit, and bun pm audit which require network access to established, well-known vulnerability databases (SKILL.md).
[PROMPT_INJECTION]: Features an indirect prompt injection surface as the skill is designed to ingest and process untrusted project files during an audit.
Ingestion points: Reads external project files using grep and git grep in SKILL.md.
Boundary markers: None present in instructions for delimiting external content.
Capability inventory: Includes the ability to execute shell commands and file system searches (SKILL.md).
Sanitization: No sanitization or escaping mechanisms are described for the data processed from the codebase.

security-audit