healthcare-admin
Pass
Audited by Gen Agent Trust Hub on Sep 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input to generate files in the workspace, which introduces a risk where malicious data could influence the generated output.
- Ingestion points: Facility context, EHR platform details, and accreditation body priorities defined in the 'Working mode' section of SKILL.md.
- Boundary markers: No specific delimiters or safety instructions are used to separate user-provided facility data from the agent's logic.
- Capability inventory: The skill requests 'workspace-write' permissions in the YAML frontmatter, allowing it to write policies, reports, and templates to the filesystem.
- Sanitization: There are no instructions for sanitizing or validating external context data before it is interpolated into survey-ready outputs.
- [METADATA_POISONING]: The skill's metadata contains a non-existent model version ('gpt-5.4'), which is a deceptive metadata practice typically used to influence model behavior or bypass version-based logic.
Audit Metadata