The Agent Skills Directory

[COMMAND_EXECUTION]: The probe.py script executes local system commands (git, gh) to gather repository information. This is implemented securely using subprocess.run with argument lists (avoiding shell execution) and a 60-second timeout to prevent resource exhaustion.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository being audited (e.g., file contents, commit messages, issue summaries). While this presents a vulnerability surface, it is mitigated by the structured processing in probe.py and the intended audit context.
Ingestion points: probe.py reads tracked files and git history in the target repository to classify SDLC stages.
Boundary markers: The skill instructions emphasize read-only operation and produce structured JSON output for the agent to interpret.
Capability inventory: The script performs local command execution (git, gh) and file system access to collect evidence.
Sanitization: A built-in secret scanner identifies and masks sensitive patterns (such as AWS keys, GitHub tokens, and private keys) in the audited content before returning results to the agent, reducing exposure risk.

sdlc-audit