prior-auth-review-skill
Pass
Audited by Gen Agent Trust Hub on Jul 30, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill processes untrusted external data from healthcare provider documentation (PDFs of clinical notes, imaging reports, and lab results). This information is extracted and used to perform medical necessity assessments and generate recommendations. The instructions do not specify boundary markers or sanitization steps for the ingested clinical data, creating a surface for indirect prompt injection where malicious instructions hidden in medical records could attempt to influence the AI's approval or denial decisions.
- Ingestion points: Processes uploaded PDF files and clinical notes in
references/01-intake-assessment.md(Step 1). - Boundary markers: No explicit boundary markers or 'ignore' instructions for clinical content are defined.
- Capability inventory: The skill performs file system writes (
waypoints/,outputs/) and network operations via MCP connectors and WebFetch for code validation. - Sanitization: No explicit sanitization or filtering of the extracted text from clinical documents is described.
Audit Metadata