search-skills

SUSPICIOUS: the skill's capabilities fit its research-summary purpose, and data flows are broadly consistent with arXiv plus an LLM backend. However, execution trust is only moderate because the guide tells the agent/user to clone and run mutable code from a personal GitHub repository with unpinned dependencies, and it allows API traffic to a custom endpoint via OPENAI_BASE_URL. This is better classified as supply-chain and endpoint-trust risk than malware.