The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/detect_secrets.py calls git ls-files using subprocess.run to determine if audited files are tracked by Git. This is a local, non-shell execution used solely for metadata enrichment and does not pose a security risk.
[DATA_EXFILTRATION]: The skill analyzes source code for secrets but operates entirely offline. It contains no network-capable libraries (e.g., requests, urllib) and does not perform any outbound requests. Discovered secrets are automatically redacted in snippets to prevent exposure.
[EXTERNAL_DOWNLOADS]: The skill is self-contained and uses only the Python standard library. It does not download or install external packages, scripts, or binaries at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code, which creates a potential surface for indirect injection. However, it is well-defended:
Ingestion points: scripts/scan.py reads file content from a user-specified target directory.
Boundary markers: SKILL.md provides explicit instructions to the agent to treat target code only as data/evidence and to disregard embedded instructions.
Capability inventory: The skill is limited to file system read access, local git checks, and standard output reporting.
Sanitization: The scanner implementation includes automated redaction of identified credentials.

crypto-secrets