Discover and explore top open-source AI tools and projects—updated daily.
KimYx0207Governed execution for AI coding assistants
Top 99.8% on SourcePulse
Summary
Meta_Kim provides a governance layer for AI coding assistants, addressing the challenge of managing complex, multi-step AI-driven development. It targets engineers and power users by structuring AI coding workflows, ensuring clarity of intent, controlled execution, verifiable results, and persistent learning, thereby transforming AI collaboration from a potential mess into a disciplined, runnable system.
How It Works
The core architecture employs an "8-stage workflow" (Critical, Fetch, Thinking, Execution, Review, Meta-Review, Verification, Evolution) acting as a "hidden skeleton." This is augmented by "dynamic dealing," a system of adaptive cards that inject flexibility based on task complexity and risk. Execution is governed by structured "contracts" (e.g., intentPacket, dispatchBoard) and "gates" that enforce pass/fail conditions. A novel three-layer memory system—local Memory, project-level Graphify (knowledge graph), and SQL (vector retrieval)—enhances continuity, reduces hallucinations, and optimizes token usage by providing agents with structured context and learned lessons.
Quick Start & Requirements
Installation is straightforward via npx --yes github:KimYx0207/Meta_Kim meta-kim or by cloning the repository, running npm install, and executing node setup.mjs. Prerequisites include Node.js (version 22.13.0 recommended for meta:sync) and optionally Python for graphify.
Highlighted Details
Maintenance & Community
The project is maintained by KimYx0207. Community interaction primarily occurs through the GitHub repository.
Licensing & Compatibility
The core Meta_Kim project is licensed under the Apache License 2.0, permitting commercial use with attribution. Optional skill repositories carry their own licenses, predominantly MIT.
Limitations & Caveats
Non-default projections (OpenClaw, Cursor) require explicit maintainer approval and self-test evidence. Candidate probes are not formally supported until adapter development and validation. The system is designed for complex, cross-file tasks, not simple single-file edits. Setup requires Node.js and potentially Python.
1 day ago
Inactive