auto-improve

SUSPICIOUS. The skill’s core purpose is coherent with local prompt/doc optimization, but its actual footprint is unusually broad: autonomous self-modification, silent conversation harvesting, durable memory/profile storage, creation of new skills, and optional shell-command injection. I found no clear external credential theft or exfiltration endpoint, so this is not confirmed malware, but it is a high-risk autonomous/persistence skill with disproportionate privacy and prompt-injection exposure.