gh:optimize

What problem does it solve?

The Skill orchestrates metric-driven optimization loops by repeatedly evaluating multiple code or configuration variants against a shared measurement harness, then surfacing a durable best solution. It emphasizes reproducibility, guardrails, and crash-safe logging to support long-running experiments.

Core Features & Use Cases

• Orchestrates iterative experiments with a repeatable harness and explicit guardrails
• Supports both objective hard metrics and LLM-as-judge scoring for qualitative targets
• Manages hypothesis backlog, per-experiment worktrees, and persistent result logs for auditability
• Suitable for tuning memory, latency, clustering quality, prompt efficiency, and other measurable outcomes

Quick Start

Provide the optimization goal or path to a spec YAML file to start the workflow.