lab-interpreter
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data extracted from medical reports (images, PDFs, or text), creating a potential vector for indirect prompt injection.
- Ingestion points: User-uploaded medical documents and clinical reports containing arbitrary text (Step 1).
- Boundary markers: None. The skill does not define specific delimiters or instructions to prevent the agent from obeying instructions embedded within the medical reports.
- Capability inventory: The skill has file-writing capabilities to
/mnt/user-data/outputs/and can modify long-term agent memory viamemory_user_edits. No network or shell tools are used. - Sanitization: Absent. The skill uses extracted data (e.g., patient name) to dynamically generate file paths without explicit validation or sanitization, potentially allowing path manipulation if a malicious name is extracted.
Audit Metadata