skill-improver

Orchestrate hypothesis-driven skill package improvements with frozen evaluators and rollback.

Updated Apr 27, 2026
One-click install
npx skills add https://github.com/ginmp8/rhapsodia --skill skill-improver-ginmp8
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-improver
Source: https://github.com/ginmp8/rhapsodia/tree/main/skills/skill-improver
Command: npx skills add https://github.com/ginmp8/rhapsodia --skill skill-improver-ginmp8

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

Inspect and improve existing skill packages through bounded experiments: freeze an evaluator, measure baseline, discover or load a hypothesis, apply a minimal patch, re-evaluate, and report with rollback safety.

Core Features & Use Cases

  • Bounded hypothesis-driven improvement: orchestrate a single patch cycle with guardrails, rollback, and measurable delta.
  • Safe evaluation workflow: freeze evaluator inputs, enforce change gates, and validate packaging before finalization.
  • Reusable infrastructure: supports discovery backlog, patch templates, and structured reporting for audit trails.

Quick Start

Provide the target skill folder and a frozen evaluator, then run the improvement loop to test a bounded hypothesis.

Frequently Asked Questions about skill-improver

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
What is hypothesis-driven skill improvement and how does it work?

Hypothesis-driven skill improvement orchestrates bounded experiments on existing skill packages by freezing an evaluator, measuring a baseline, applying a minimal patch, and re-evaluating to validate changes with rollback safety.

How do I safely apply patches to an existing skill package?

To safely apply patches to a skill package, provide the target skill folder and a frozen evaluator to run the improvement loop, which enforces change gates, validates packaging, and ensures rollback safety before finalization.

Can I audit and benchmark skill packages without risking breaking changes?

Yes, you can audit and benchmark skill packages safely by freezing evaluator inputs and enforcing change gates, which validate minimal patches and provide a descriptive patch record for rollback if needed.

What do I need to run a skill improvement loop?

You need a target skill folder and a frozen evaluator to run the skill improvement loop, which then measures the baseline, discovers or loads a hypothesis, and applies a minimal patch for validation.

Does the skill improvement process support rollback for failed validation?

Yes, the skill improvement process supports rollback for failed validation by enforcing change gates and validating packaging before finalization, ensuring safe reversion if the evaluator rejects the patch.

How does a frozen evaluator ensure safe skill validation?

A frozen evaluator ensures safe skill validation by locking inputs to measure the baseline and re-evaluate minimal patches consistently, preventing evaluator drift during the bounded improvement experiment.