autofix

SUSPICIOUS: the stated purpose is coherent, but the skill’s main mechanism is high-risk because it treats external PR comment content as executable agent instructions and can autonomously edit code, commit, push, and post comments. The `bkt` dependency appears openly published and somewhat verifiable, so the main concern is indirect prompt injection and automated repository actions, not confirmed malware.