context-engineering-collection

SUSPICIOUS: the skill is coherent as hosted-agent infrastructure guidance, not overt malware, but it normalizes high-impact autonomous behavior, sensitive GitHub token handling, and broad remote execution. Risk is driven by scope and execution model rather than deceptive data exfiltration; the OpenCode curl|bash reference is a secondary supply-chain concern, not the main issue.

SUSPICIOUS: The stated purpose is coherent for an agent-debugging skill, and the only evidenced API flow aligns with official MiniMax endpoints. However, the undocumented and unverifiable `rto` CLI is a major install-trust gap, and the skill handles highly sensitive reasoning/session traces with optional prompt mutation and skill generation. Main concern is execution provenance and trace exposure, not confirmed malware.