autoresearch

Pass

Audited by Gen Agent Trust Hub on May 15, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides instructions for an AI agent to act as a research architect, focusing on autonomous optimization loops. It provides a structured output template for technical design and business modeling.
  • [EXTERNAL_DOWNLOADS]: The skill mentions cloning github.com/karpathy/autoresearch. This is a reference to a repository from a well-known researcher in the AI field and is considered a safe reference in this context.
  • [PROMPT_INJECTION]: The skill exhibits a surface for indirect prompt injection (Category 8) as it processes user-provided goals to generate technical plans and commands.
  • Ingestion points: The user provides a "goal or domain" which is processed by the agent.
  • Boundary markers: None are specified to isolate the user-provided goal from the system instructions.
  • Capability inventory: The agent is instructed to produce "technical setup plans" and "exact shell commands" (SKILL.md).
  • Sanitization: No sanitization or validation of the user input is described in the instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
May 15, 2026, 05:58 PM
Security Audit — agent-trust-hub — autoresearch