The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses highly sensitive file paths including .env, database/auth_db.py, and git history to perform credential hygiene checks and secret scanning. While it utilizes hashing for reporting, the access to raw secrets in these files is a high-privilege operation.
[COMMAND_EXECUTION]: The audit script executes several external command-line tools via subprocess calls, including git, uv, npm, bandit, detect-secrets, and pip-audit, to assess the repository's security posture.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted codebase content and presents results to the agent, creating a surface where malicious instructions embedded in source files could influence the agent's behavior. 1. Ingestion points: Reads all project source files (.py, .ts, .tsx), environment files, and git logs. 2. Boundary markers: No delimiters or ignore instructions are used to isolate audited content from agent instructions. 3. Capability inventory: Executes system commands and writes filesystem reports. 4. Sanitization: Findings derived from untrusted file content are displayed without sanitization.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to the official npm and PyPI registries via npm-audit and pip-audit to verify dependency security. These are well-known services required for the skill's purpose.

security-audit