The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to execute local git commands (e.g., git rev-parse, git status, git diff) to gather repository metadata and verify file integrity. These operations are essential for the skill's auditing functionality and are restricted to the target repository.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface. The audit_repo.py script reads content from untrusted files within the target repository (such as AGENTS.md and various code files) to generate a summary report. Malicious instructions embedded in the audited repository could theoretically influence an AI agent that subsequently processes the generated report.md or findings.json files.
Ingestion points: Content is read from AGENTS.md, .env, and other repository files in audit_repo.py (e.g., in _read_text_best_effort and scan_possible_secrets).
Boundary markers: The output reports do not use specific delimiters or warnings to isolate content ingested from the target repository.
Capability inventory: audit_repo.py can read files and write to a user-defined output directory. remediate_repo.py can write template files to the repository.
Sanitization: The scripts use regex to extract information but do not explicitly sanitize content before including it in the generated reports.

repo-compliance-audit