ralph

SUSPICIOUS rather than malicious. The documented purpose matches autonomous coding, and there is no explicit credential theft or exfiltration path in the provided text, but the skill's main capability is delegated to an unreviewed local shell script and an unspecified secondary skill. Because the executable footprint is not verifiable from the skill content, the overall risk is high even without confirmed malware.

The script appears intended to automate an AI coding-agent loop. It contains a concrete archive path traversal risk because branch names are used to construct filesystem paths without validation. It also launches an external agent with permissions skipping enabled, which can permit consequential actions depending on the agent and prompt. No direct malware or data theft is evident in this file.