3 ms·
Show HN: Built a small CLI for self-improving OpenClaw agent loops
It reviews a plan or output with 1–3 LLMs and returns: PASS / REVISE / BLOCK
Use case:
- review a plan before execution
- review an output before sending or relying on it
- loop until the result is good enough to proceed
Very small repo (255 lines of code)