evidence-level-ranker
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external literature, which is an ingestion point for untrusted data. While an attack surface exists, the risk is minimal due to the skill's restricted capabilities.
- Ingestion points: User-supplied research papers and abstracts (SKILL.md).
- Boundary markers: None identified.
- Capability inventory: Restricted to text generation; no access to network, files, or shell tools.
- Sanitization: No explicit input sanitization rules.
- [SAFE]: The skill implements significant defensive measures and integrity checks.
- Integrity Rules: Hard Rules 11-14 strictly prohibit fabricating citations, identifiers, or study findings.
- Scope Enforcement: The skill explicitly prohibits providing medical advice or acting as a treatment recommendation engine.
- Trusted Source: References to the author's repository (github.com/aipoch/medical-research-skills) are handled neutrally as vendor-owned resources.
Audit Metadata