scientific-critical-thinking
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill represents an ingestion surface for untrusted scientific data, such as research papers or claims (SKILL.md). No explicit boundary markers or delimiters are provided to separate data from instructions. However, the skill's capabilities are restricted to standard file tools and specific vendor scripts, and its analytical nature encourages scrutiny of input.\n- [COMMAND_EXECUTION]: The skill provides instructions for the agent to execute a schematic generation script (generate_schematic.py) from a related vendor repository. This is a functional requirement for creating visual aids.\n- [EXTERNAL_DOWNLOADS]: The skill directs the agent to fetch citation metadata from the arXiv API (export.arxiv.org), a well-known academic repository. This operation is used for standard bibliography management.
Audit Metadata