Stealthy-McStealth
OrganizationMake your Claude Code agent work like a real harness (Hermes). Self-evolving skills + drift prevention. Research-backed.
Categories
Indexed Skills (2)
improve-urself
Extract learnings from the current conversation and encode them as persistent skills. Trigger when the user types /improve-urself (with optional topic args like "/improve-urself json formatting" or "/improve-urself tool-use bash"), or says things like "save that as a skill", "we should remember this", "turn that into a rule", "learn from what just happened", "add this to your skills", or "don't make that mistake again". Also trigger proactively after a long conversation where multiple corrections were made. Always use this skill — don't just summarize the learning in chat and move on.
reload-context
Re-read and re-anchor to all loaded skills and active instructions. Use when the user types /reload-context or /brainwash, or says things like "start fresh", "forget what you think you know", "re-read the rules", "you're drifting", "you stopped following the instructions", "go back to basics", or "reset". Also trigger proactively if you notice yourself improvising around a loaded skill's instructions rather than following them — that's the drift this skill is designed to fix.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.