diagnose

SUSPICIOUS. The skill's purpose broadly matches troubleshooting, and its local file reads are proportionate to Claude Code diagnostics. However, it instructs the agent to silently fetch and execute a remote shell script from a mutable GitHub branch with no pinning or verification, creating a significant supply-chain risk. Data flows remain mostly aligned with the stated purpose, so this is better classified as high-risk/vulnerable rather than confirmed malware.