The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill contains multiple auditing tools designed to scan a project's codebase and documentation for compliance gaps (e.g., gdpr_compliance_checker.py, hipaa_risk_assessment.py, and qsr_compliance_checker.py). This functionality creates a surface for indirect prompt injection.
Ingestion points: The scripts recursively scan the project directory using Path.rglob("*") and read the content of all files matching relevant extensions (.py, .js, .md, .json, etc.) to search for data patterns and compliance keywords.
Boundary markers: The output reports generated by these tools do not currently implement specific boundary markers or instructions to the agent to disregard potential instructions found within the audited content.
Capability inventory: The analysis confirmed that these scripts lack dangerous capabilities; they do not perform network requests, spawn subprocesses, or use dynamic execution functions like eval() or exec().
Sanitization: Scanned content is processed via regex and included in results without additional sanitization, relying on the agent's own safety guardrails.
Note: This risk is considered low and is directly associated with the primary, legitimate purpose of the skill.
[SAFE]: A comprehensive review of the entire collection found that all scripts are transparent and restricted to the Python standard library. No suspicious network activities, hardcoded credentials, or persistence mechanisms were found. The skill maintains a clear separation between its internal tracking data and the user's project data.

ra-qm-skills