litreview
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill fetches academic paper metadata from well-known services, including the National Center for Biotechnology Information (NCBI) PubMed E-utilities and the OpenAlex API. These are established, legitimate sources for scholarly information.
- [COMMAND_EXECUTION]: The agent uses shell capabilities to execute local Python scripts included with the skill. These scripts handle search queries, deduplicate papers, and maintain an audit log of the research session in a local directory (
~/.litreview_sessions). - [INDIRECT_PROMPT_INJECTION]: The skill ingests external data from academic databases (such as paper titles, author names, and journal venues) to produce its final orientation guide.
- Ingestion points: Paper metadata returned by the PubMed and OpenAlex APIs as processed in
scripts/free_search.py. - Boundary markers: While the skill uses structured logic to synthesize results, it does not explicitly delimit all interpolated data fields within the generated guide.
- Capability inventory: The skill has capabilities for network search, internal script execution via
bash_tool, and file system writes for session logs and DOCX generation. - Sanitization: The skill utilizes standard URL encoding for outbound API queries and processes data through dedicated aggregation scripts before final interpolation into the research report.
Audit Metadata