learning-quality

Capture and generalize actionable learnings from AI agent operations.

102|10|Updated Mar 4, 2026
One-click install
npx skills add https://github.com/closedloop-ai/claude-plugins --skill learning-quality
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: learning-quality
Source: https://github.com/closedloop-ai/claude-plugins/tree/main/plugins/self-learning/skills/learning-quality
Command: npx skills add https://github.com/closedloop-ai/claude-plugins --skill learning-quality

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill provides a structured process to identify, generalize, and capture valuable learnings during AI agent runs, preventing repeated mistakes and building organizational knowledge.

Core Features & Use Cases

  • Decision Tree: Guides users on whether a learning is worth capturing based on criteria like non-obviousness, generalizability, and scope.
  • Scope Classification: Differentiates between project-specific learnings (CLAUDE.md) and globally applicable patterns (org-patterns.toon).
  • Pattern Generalization: Provides examples and a formula for transforming specific instances into reusable principles.
  • Conflict Resolution: Outlines a process for checking and resolving contradictions with existing learned patterns.
  • Capture Workflow: Details the steps for writing and saving learnings in the correct format and location.
  • Use Case: An AI agent repeatedly makes a specific coding error due to a misunderstanding of a framework's enum usage. This skill ensures that misunderstanding is captured as a generalizable pattern, preventing future agents from making the same mistake.

Quick Start

Use the learning-quality skill to capture a new pattern about handling optional dictionaries in FastAPI.

Frequently Asked Questions about learning-quality

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

FAQPage Schema
How do I capture generalizable learnings from AI agent operations?

Capture AI agent learnings by applying a structured decision tree to evaluate non-obvious patterns for reusability and scope, resolving any conflicts before saving the generalized patterns in JSON format to prevent repeated mistakes.

What is the best way to generalize specific coding errors into reusable AI patterns?

Generalize specific coding errors into reusable AI patterns by applying a transformation formula that abstracts the instance into a principle, verifying non-obviousness and generalizability before saving the learning.

How do I know if an AI agent learning belongs in CLAUDE.md or org-patterns.toon?

Classify the learning's scope to determine its destination: project-specific learnings go to CLAUDE.md, while globally applicable patterns across projects go to org-patterns.toon.

How do I resolve conflicts with existing learned patterns before saving new agent knowledge?

Resolve conflicts with existing learned patterns by executing a checking process that validates new AI agent learnings against current saved patterns to identify and address contradictions before finalizing the JSON capture.

When should I not capture an AI agent learning into the knowledge base?

Avoid capturing an AI agent learning if it fails the decision tree criteria for non-obviousness, lacks generalizability across projects, or is too narrow in scope to improve future agent performance.