national-science-review
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed to perform text analysis and provide recommendations based on heuristics for scientific publishing. It does not contain executable code, scripts, or network exfiltration patterns. The stated purpose of the skill is consistent with its implementation.\n- [PROMPT_INJECTION]: The skill operates on user-supplied manuscripts, creating an attack surface for indirect prompt injection (Category 8). The risk is assessed as safe because the skill has a limited capability set and produces output constrained to a specific text template, preventing the execution of embedded instructions.\n- [DATA_EXFILTRATION]: The instructions include references to local resource files (
../../resources/source-basis.mdand../../resources/official-source-map.md). These references are used for information verification and grounding against official sources and do not involve sensitive system files or credentials.
Audit Metadata