self-healing

SUSPICIOUS: the skill is broadly aligned with automated debugging and repair, and it does not show credential theft or off-platform exfiltration. However, it grants an AI agent autonomous write/commit authority and executes project scripts based on external failure content, creating meaningful operational and prompt-injection risk disproportionate to a normal diagnostic helper.