self-improving-agents

ThumbGate Pre-Action Checks self-improve from ranked lessons and repeated failures, hard-block detected secret leaks, and block matches in strict mode.

From self-improving agents to co-evolving, open-ended agent ecosystems.

Turn recurring Claude Code and OpenAI Codex mistakes into durable, reviewable workspace memory with safe local context patches.

Karpathy-inspired LLM wiki and Google WikiSkill implementation for grounded agent memory, persistent knowledge and evaluation-gated self-improving Agent Skills.

Turn your AI agent into a capability amplifier — it watches your real work, finds where you're actually stuck, traces it to the discipline that studied it for decades, and feeds what it learns into your mneme. Pairs with AgentGameLab/mneme.

Your coding agent should never make the same mistake twice. Turn failed attempts into verified, reusable skills.