create-skill-autoresearch
Pass
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill manages sensitive API keys using environment variables (such as
JUDGE_API_KEY) rather than hardcoding credentials, which aligns with secure development practices.\n- [SAFE]: Network activity is restricted to interacting with user-configured, OpenAI-compatible API endpoints for the purpose of LLM-as-judge evaluation and benchmarking.\n- [SAFE]: Dynamic code generation for evaluation scripts (evaluate.sh) employs Python'sjsonmodule to safely encode file contents, effectively preventing injection into the request structure.\n- [SAFE]: External repository references point to the author's own GitHub organization (a-tokyo), representing legitimate vendor-owned resources for the skill's operation.
Audit Metadata