scienceworld-reset

Pass

Audited by Gen Agent Trust Hub on Feb 20, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE] (SAFE): No malicious patterns, prompt injections, or obfuscation techniques were detected. The tool's logic is dedicated to simulation environment management and adheres to safe coding practices.
  • [EXTERNAL_DOWNLOADS] (SAFE): The tool uses the 'scienceworld' Python package, which is a recognized library for reinforcement learning and scientific simulations. No dynamic script downloads or remote code execution patterns were found.
  • [DATA_EXFILTRATION] (SAFE): The tool does not access sensitive file paths or credentials. Data handling is restricted to simulation-related metadata and environment observations, with no unauthorized network transmission.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 20, 2026, 10:51 PM