Self-evolving knowledge engine. Learn from any content, process it through 8-phase SEED protocol, track progression like an MMORPG, export knowledge as portable skill files. Teaches agents to research before building.
The SEED protocol — teaches your agent to learn how to learn. Auto-evolving agent skill that combines persistent memory, autonomous skill extraction, multi-agent emergence, and voice interaction into one protocol. Works with Claude Code, OpenClaw, Cursor, Codex. Use when: (1) Starting ANY session — SEED auto-activates to evaluate learnings, (2) Complex tasks requiring multi-perspective analysis (8 Owls emergence), (3) After completing tasks — auto-extracts reusable skills from discoveries, (4) When agent needs persistent context across sessions or devices, (5) When orchestrating multiple agents (conductor mode), (6) User says "evolve", "learn", "remember", "teach me", or "what do you know?" Trigger conditions: Every session start, every task completion, every non-obvious discovery. This skill should be evaluated for relevance on EVERY interaction.