The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill set contains several scripts designed to recursively scan a user-specified project directory to detect GDPR violations, HIPAA risks, or QSR gaps. This creates a surface where the agent could ingest malicious instructions hidden in the files being analyzed.
Ingestion points: Scripts such as gdpr_compliance_checker.py, qsr_compliance_checker.py, and hipaa_risk_assessment.py utilize Path.rglob and read_text to ingest content from any file within the provided project directory.
Boundary markers: The findings from these scans are returned to the agent context as raw text or structured JSON without delimiters or explicit instructions for the LLM to ignore potentially adversarial content embedded in the target files.
Capability inventory: The tools perform recursive file system reads across various extensions (.py, .js, .md, etc.) and present the discovered snippets to the agent. This content could include maliciously crafted comments or strings intended to hijack the conversation.
Sanitization: There is no evidence of filtering or sanitization being applied to the ingested file content before it is reported back to the agent context.

ra-qm-skills