p-article-evaluated
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied academic manuscripts, which constitutes a surface for indirect prompt injection. If a manuscript contains hidden instructions intended to override the agent's behavior, the agent could potentially be influenced. However, the skill implements a rigid, step-by-step diagnostic framework (Fact Table, Reverse Outline, Dimension Rating) which significantly reduces the impact of such injections compared to general-purpose chat agents. Additionally, the skill lacks capabilities for network operations or persistent file modifications, further limiting the risk profile.
Audit Metadata