The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes environment-local security tools including pip-audit, npm audit, cargo audit, and govulncheck to gather deterministic vulnerability data from the target project. These are standard developer tools used for their intended purpose.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because its core function involves reading and analyzing untrusted codebase content. Maliciously crafted code or comments in the audited project could attempt to influence the agent's analysis.
Ingestion points: Reads all source files and configuration metadata within the user-specified {TARGET_DIR}.
Boundary markers: Employs a 'Unified Output Contract' that forces agents to provide findings in a structured Markdown table, limiting the potential for open-ended injection effects.
Capability inventory: Has the ability to execute shell commands for auditing (Phase 0) and write structured reports to a local .audit/ directory.
Sanitization: Uses a 'Phase 1.5' verification step where secondary agents re-evaluate high-severity claims to filter out false positives and potential adversarial content.

codebase-audit