Mulch Self Improver

Let your agents grow 🌱

~27.5%

fewer characters (token proxy) vs baseline — same session + troubleshooting + style & memory. 1441 chars (~352 tokens) saved per full mix.

Same tasks, less context. Proven side-by-side against Self Improving Agent — Rank #2 on ClawHub.

~14% session ~54% troubleshooting ~33% style & memory
Full benchmark comparison BENCHMARK.md

Baseline: Self Improving Agent — Rank #2 on ClawHub. Built with the visual-explainer skill.