literature-review
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for ingesting and processing data from external academic databases such as PubMed, arXiv, and Semantic Scholar.\n
- Ingestion points: Academic abstracts, full-text papers, and technical reports (SKILL.md).\n
- Boundary markers: The skill recommends using structured extraction tables to organize data, which helps separate external content from the synthesis logic, though explicit instruction-isolation markers are not defined.\n
- Capability inventory: The skill is entirely documentation-based and does not provide or call any shell scripts, network tools, or file-writing utilities.\n
- Sanitization: No explicit sanitization or filtering of external paper content is described.\n- [NO_CODE]: The skill is purely comprised of instructional markdown content and contains no executable code, dependencies, or configuration for dynamic execution.
Audit Metadata