clinical-reports-hardened
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the Bash tool to run local Python scripts (e.g.,
scripts/generate_schematic.py) for medical validation and diagram generation. It contains explicit instructions for the agent to quote and escape clinical data to prevent shell injection. - [DATA_EXFILTRATION]: The skill handles sensitive Protected Health Information (PHI) but includes strict defensive instructions prohibiting the transmission of PHI to external network endpoints, APIs, or shared drives via shell commands.
- [PROMPT_INJECTION]: The skill ingests clinical data and patient history, presenting a surface for indirect prompt injection.
- Ingestion points: Clinical records, patient-supplied symptoms, and history files (e.g.,
data/med_list.txt). - Boundary markers: No specific delimiters are enforced in the prompt, but the skill mandates treating all clinical data as untrusted.
- Capability inventory: The agent has access to Bash, Write, and Edit tools.
- Sanitization: The instructions require the agent to escape shell arguments and scan for 18 HIPAA identifiers to ensure de-identification.
Audit Metadata