harness-improvement

Integrate session learnings and user requests into AI harnesses across Claude, Codex, Cursor, and OpenCode.

Updated Jun 24, 2026

Search Tags:#continuous improvement #agent management #skill update #cross-platform adaptation #AI harness

Ready to run

npx skills add https://github.com/AvatarGanymede/pcraft --skill harness-improvement

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: harness-improvement
Source: https://github.com/AvatarGanymede/pcraft/tree/main/.agents/skills/harness-improvement
Command: npx skills add https://github.com/AvatarGanymede/pcraft --skill harness-improvement

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

This Skill addresses the need to incorporate learnings from AI sessions and explicit user requests into the AI harness, ensuring continuous improvement and adaptability.

Core Features & Use Cases

Session Learning Integration: Incorporate recurring learnings from AI sessions into the harness.
Skill and Agent Updates: Update or create skills, agents, subagents, commands, and guidance files.
Cross-Platform Adaptation: Adapt harness files across Claude, Codex, Cursor, and OpenCode platforms.
Use Case: When a user requests to update a skill based on a specific task's outcome or to record a new learning from an AI session.

Quick Start

Run the 'harness-improvement' skill to update a skill based on learnings from a recent AI session.

Frequently Asked Questions about harness-improvement

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

FAQPage Schema

How do I integrate recurring learnings from AI sessions into my harness?▼

You can integrate session learnings by applying the harness-improvement skill, which captures recurring outcomes and explicit requests to update your AI harness. This ensures continuous improvement by modifying guidance files and agent structures based on what was learned during previous AI interactions.

Can I update skills and agents across Claude, Codex, Cursor, and OpenCode platforms?▼

Yes, you can update skills, agents, and commands across Claude, Codex, Cursor, and OpenCode platforms. The harness-improvement skill performs cross-platform adaptation, ensuring your AI workflow updates and guidance files are applied consistently across all four supported environments.

What's the best way to record a new learning from an AI session into my workflow?▼

The best way to record a new learning is to run the harness-improvement skill, which captures session outcomes and explicit requests to update your harness structure. This updates your skills and guidance files automatically, ensuring continuous improvement of your AI workflows without manual file editing.

Do I need knowledge of harness structure to update AI skills based on user requests?▼

Yes, you need knowledge of harness structure and user requests to effectively use the harness-improvement skill. Understanding how skills, agents, and guidance files are organized ensures that explicit requests and learnings are correctly integrated into your AI workflow for successful continuous improvement and cross-platform adaptation.

Why does cross-platform harness adaptation matter for AI workflow updates?▼

Cross-platform harness adaptation ensures AI workflow updates and skill modifications are consistently applied across Claude, Codex, Cursor, and OpenCode. Without it, learnings and explicit requests would need manual translation for each platform, breaking the continuous improvement cycle and fragmenting agent management.