idea-creator

SUSPICIOUS: the stated purpose is legitimate research ideation, and external model use appears to be official OpenAI infrastructure, but the skill is overpowered for that purpose. Its combination of broad Bash access, Agent delegation, and ingestion of untrusted web/paper content creates a high prompt-injection and unintended execution risk, with moderate autonomous compute-action risk. No clear credential harvesting, covert exfiltration, or confirmed malware behavior is present.