kimchi:halmoni

Analyze execution outcomes and user feedback to propose and apply versioned AI skill improvements.

8|Updated Jan 31, 2026

Search Tags:#planning #validation #self-improvement #ai #skills #personas

Ready to run

npx skills add https://github.com/Tromml/kimchi --skill kimchi-halmoni

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: kimchi:halmoni
Source: https://github.com/Tromml/kimchi/tree/main/plugins/kimchi/skills/halmoni
Command: npx skills add https://github.com/Tromml/kimchi --skill kimchi-halmoni

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve?

This Skill enhances the AI's planning capabilities by learning from past execution outcomes and user feedback, leading to more accurate and efficient task specifications.

Core Features & Use Cases

Self-Improvement: Observes execution results and user feedback to propose and apply incremental improvements to skills, validators, and personas.
Versioned Updates: Automatically updates skills and logs changes, ensuring a traceable improvement history.
Use Case: After a planning cycle produces a suboptimal plan, Halmoni can analyze the execution logs, identify where the AI struggled, and suggest specific adjustments to the planning skills or personas to prevent similar issues in the future.

Quick Start

Run the Halmoni self-improvement system to analyze the execution of bead ID 008.

Frequently Asked Questions about kimchi:halmoni

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

FAQPage Schema

How does AI self-improvement for planning work?▼

AI self-improvement for planning works by analyzing past execution outcomes and user feedback to propose, validate, and apply incremental enhancements to AI skills, validators, and personas. This iterative refinement ensures planning logic becomes more accurate over time.

How do I make my AI planning logic learn from past execution outcomes?▼

You make AI planning logic learn from past execution outcomes by running a self-improvement system that analyzes execution logs to identify where the AI struggled. It then suggests specific adjustments to planning skills or personas to prevent similar issues in the future.

Can I version and log incremental enhancements applied to AI validators and personas?▼

Yes, you can version and log incremental enhancements applied to AI validators, skills, and personas. The system automatically updates skills and logs changes, ensuring a traceable improvement history for all applied modifications.

What is the best way to use user feedback for AI self-improvement?▼

The best way to use user feedback for AI self-improvement is to feed it directly into a validation system that observes both the feedback and execution results. The system then proposes and applies validated enhancements to refine execution logic.

Do I need execution logs to propose and validate adjustments to AI personas?▼

Yes, you need execution logs or direct user feedback to propose and validate adjustments to AI personas. The system analyzes these inputs to identify specific areas where the AI struggled during a planning cycle to suggest targeted improvements.

Why does AI planning produce suboptimal plans after a planning cycle?▼

AI planning produces suboptimal plans when execution logic struggles with specific task constraints. Analyzing execution logs from the failed cycle allows a self-improvement system to identify these struggle points and apply targeted skill adjustments to prevent future issues.