scientific-critical-thinking
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to evaluate external scientific claims and research methodology, creating an attack surface where untrusted input is ingested into the agent context. It lacks explicit instructions for data sanitization or the use of boundary markers for the ingested text. The skill's capabilities to read, write, and edit files, combined with its shell command execution facility, characterize this as a typical indirect prompt injection surface.
- [COMMAND_EXECUTION]: The documentation provides an example command to execute a local Python script (
generate_schematic.py) belonging to the author's related schematics skill. - [EXTERNAL_DOWNLOADS]: The skill instructs the agent to perform network operations, including fetching citation metadata from the ArXiv API and transmitting prompt data to the OpenRouter API for visual aid generation. These are legitimate interactions with well-known scientific and technological services.
Audit Metadata