figure-first-paper-reader
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external documents such as PDFs and research figures, creating an attack surface for indirect prompt injection. This risk is effectively mitigated by the skill's lack of privileged capabilities (no network access, file system write, or command execution) and its reliance on a strict 15-rule integrity framework. These rules, combined with a mandatory 8-step execution process, ensure the agent distinguishes between observed evidence and narrative interpretation, preventing untrusted data from overriding the skill's logic.
- [SAFE]: The skill's reference modules and core instructions are focused solely on evidentiary auditing. It references its development repository on GitHub (aipoch/medical-research-skills) and exhibits no signs of obfuscation, persistence mechanisms, or privilege escalation. The skill demonstrates best practices for medical research analysis by explicitly prohibiting the fabrication of bibliographic metadata like PMIDs or DOIs.
Audit Metadata