scientist-run
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions to execute shell commands, specifically listing directories using
lsand running a Python scriptlaunch_scientist.pylocated in a relative vendor path. - [INDIRECT_PROMPT_INJECTION]: The skill implements an autonomous scientific discovery loop which ingests templates and generates research content, creating a vulnerability surface where untrusted data could influence agent behavior.
- Ingestion points: Research templates located in
../vendor/AI-Scientist/templates/and model-generated scientific ideas. - Boundary markers: None explicitly defined in the instructions to separate template data from the agent's core instructions.
- Capability inventory: File system listing (
ls) and Python script execution (python). - Sanitization: No sanitization or validation of the template content is specified within the skill instructions.
Audit Metadata