minions-autoresearch

What problem does it solve?

Automates the optimization loop for Minions skills, turning manual tuning into autonomous, test-driven improvements driven by binary evaluations.

Core Features & Use Cases

• Autonomous experimentation: performs mutations and binary evaluations with keep-or-discard decisions.
• Baseline and metrics: records baseline pass rate and tracks improvements across scenarios.
• One-mutation-per-experiment: ensures causality and prevents compound mutations.
• Dashboard and learnings: exports results to TSV/JSON, dashboards, and stores learnings for kept mutations.
• Use Case: When you want to systematically optimize a skill (e.g., minions-remember) to reach a target performance.

Quick Start

Run the Autoresearch workflow to iteratively mutate and evaluate a target skill until the target pass rate is reached or the budget is exhausted.