skill-evolution

Propose safe skill library patches from run retrospectives.

339|58|Updated Jan 10, 2026
One-click install
npx skills add https://github.com/Heyvhuang/ship-faster --skill skill-evolution-heyvhuang
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skill-evolution
Source: https://github.com/Heyvhuang/ship-faster/tree/main/skills/skill-evolution
Command: npx skills add https://github.com/Heyvhuang/ship-faster --skill skill-evolution-heyvhuang

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill enables safe, repeatable evolution of the Ship Faster skills library by transforming run retrospectives into patch recommendations rather than automatic edits, preserving human oversight.

Core Features & Use Cases

  • Retrospective-driven evolution: collects context, failures, and session data to inform skill improvements without changing the source automatically.
  • Patch proposal workflow: uses the skill-improver to generate minimal, reviewable patch candidates for manual application.
  • Artifact generation & auditing: stores evolution artifacts (candidates, review notes, run state) under runs/evolution for traceability and replay.
  • Use Case: when a big task reveals gaps or recurring failures in skills, run the evolution flow to propose targeted improvements and have a human review before applying.

Quick Start

Install the evolution skill by placing its directory into your agent's skill folder, then configure the global hooks so that end-of-task retrospectives are captured. For example:

  • Copy the skill directory to the Claude/agents skills path, e.g. mkdir -p ~/.claude/skills && cp -r skills/skill-evolution ~/.claude/skills/
  • Ensure the hooks (pre-tool, post-tool, and session-end) are enabled in your agent configuration so evolution artifacts are generated automatically.

Frequently Asked Questions about skill-evolution

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

FAQPage Schema
How do I propose patches to improve AI agent skills based on run retrospectives?

To propose patches from run retrospectives, the skill-evolution flow captures context, failures, and session summaries to generate minimal, reviewable patch candidates for manual application without auto-editing the source.

What is the best way to evolve a skill library safely without automatic source edits?

The best way to evolve skills safely is using a patch proposal workflow that transforms session data into reviewable patch candidates, preserving human oversight by requiring manual application before any changes take effect.

Do I need the skill-improver tool to generate skill evolution patches?

Yes, you need the skill-improver tool for patch generation, as it processes the captured run retrospectives and session artifacts to produce the targeted patch candidates for your skill library.

How does skill evolution store audit artifacts for traceability?

Skill evolution stores generated artifacts, including patch candidates, review notes, and run state, under the runs/evolution directory to ensure full traceability and replayability of the improvement process.

Can I use hooks to automatically capture session data for skill improvement?

Yes, you can configure global hooks like pre-tool, post-tool, and session-end in your agent configuration to automatically capture end-of-task retrospectives and generate evolution artifacts.

When should I run a skill evolution flow instead of manually editing agent skills?

You should run the skill evolution flow when a big task reveals gaps or recurring failures in skills, allowing the system to propose targeted improvements for human review rather than making immediate manual edits.