ai-scientist-v2-guide

Pass

Audited by Gen Agent Trust Hub on Mar 31, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The guide includes instructions to clone the AI-Scientist-v2 repository from Sakana AI's official GitHub repository and install dependencies via standard package managers (pip).
  • [COMMAND_EXECUTION]: Provides instructions for installing system-level dependencies like LaTeX using sudo apt-get or brew, which are necessary for the tool's scientific paper compilation feature.
  • [COMMAND_EXECUTION]: The tool's core logic involves the automated generation and execution of code modifications for machine learning experiments, which is the primary intended purpose of the research pipeline.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection as it ingests untrusted external data (PDF research papers) to influence ideation and analysis.
  • Ingestion points: External PDFs are loaded via the seed_papers parameter in IdeaGenerator (SKILL.md).
  • Boundary markers: None explicitly defined in the provided examples.
  • Capability inventory: The system performs file writes (PDF compilation) and executes experimental code modifications (Python) based on the analyzed content (SKILL.md).
  • Sanitization: No explicit sanitization or validation of the paper content is described in the guide.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 31, 2026, 10:16 PM
Security Audit — agent-trust-hub — ai-scientist-v2-guide