execute-tasks

The skill is coherent with its stated purpose as an autonomous implementation workflow, but it grants a coding agent broad local write/exec authority and lets repository-controlled text influence those actions. No clear credential theft, exfiltration, or suspicious third-party routing appears, so this is not malicious; however, the autonomous loop, git commits, and prompt-injection exposure make it a medium-risk skill.