Learning Opportunities: Deliberate Skill Development for AI-Assisted Coding

⬅️ Back to Tools

Learning Opportunities: Build Expertise While You Code

Learning Opportunities is an open-source skill by Dr. Cat Hicks that brings evidence-based learning science into your AI coding workflow. After significant architectural work (new files, schema changes, refactors), your coding agent offers optional 10-15 minute learning exercises grounded in research on how people actually build expertise.

Works with two platforms: Claude Code and Codex. Companion plugins available for automatic post-commit prompting and repo orientation.

Why This Exists

AI coding tools move fast. They generate code instantly, handle complex patterns, and keep you in a productive flow. But this speed comes with hidden costs that learning science can predict:

  • Generation effect: You accept generated code instead of generating your own, skipping the active processing that builds understanding
  • Fluency illusion: Clean code looks understood even when it isnt
  • Spacing effect: Long sessions without reflection or spacing hurt long-term retention
  • Metacognition gaps: Fast workflows leave no room to monitor what you actually know
  • Retrieval atrophy: Getting full answers means fewer opportunities to test and retrieve knowledge

Learning Opportunities counters these risks by reintroducing deliberate pauses, active generation, retrieval practice, and explicit metacognition into your coding workflow. It doesnt interrupt you randomly but waits for natural breakpoints after meaningful work.

Key Features

Evidence-Based Exercise Types

Six exercise formats grounded in well-established learning science research:

  • Prediction -> Observation -> Reflection: What do you expect to happen? Now lets see. What surprised you?
  • Generation -> Comparison: Sketch your approach before seeing the implementation
  • Trace the Path: Walk through execution step by step, predicting each transition
  • Debug This: What would go wrong here, and why?
  • Teach It Back: Explain a component as if onboarding a new developer
  • Retrieval Check-In: At the start of a session, what do you remember from last time?

Adaptive Learning Design

Exercises adjust to your work context. The skill uses a dynamic textbook approach, offering semi-worked examples drawn from your own project code. Topics come from what you just built, making each exercise relevant to real work.

Non-Intrusive Prompting

Claude asks once: “Would you like to do a quick learning exercise on [topic]? About 10-15 minutes.” Accept and it runs you through an interactive exercise. Decline and it suppresses further suggestions for the rest of the session. Capped at 2 exercises per session.

Automatic Prompting Hook (Optional)

Install learning-opportunities-auto alongside the core skill to have Claude automatically consider offering an exercise after each git commit. Works on Linux and macOS, with some setup required for Windows.

Companion Tooling

  • orient: Generate a repo orientation guide with suggested lessons. Uses strategies from empirical research on program comprehension and codebase navigation
  • learning-goal: A companion skill for semi-structured, interactive learning goal-setting using Mental Contrasting with Implementation Intentions (MCII)

Team Measurement Playbook

Comes with MEASURE-THIS.md, a lightweight playbook for running team experiments with the skill. Includes validated survey items from peer-reviewed research on developer thriving and AI skill threat, plus guidance on interpreting results and reporting to leadership.

Research-Backed Design

Unlike most AI coding tools, every design choice in Learning Opportunities is explicitly grounded in empirical research. The exercises draw from well-established findings in cognitive and educational psychology, and the design is informed by qualitative interviews with developers about their concerns with agentic coding.

Installation

Claude Code (Plugin Marketplace) 

/plugin marketplace add https://github.com/DrCatHicks/learning-opportunities.git
/plugin install learning-opportunities@learning-opportunities

Restart Claude Code to activate.

Codex (Plugin Marketplace) 

codex plugin marketplace add https://github.com/DrCatHicks/learning-opportunities.git

For local development from a checkout:

codex plugin marketplace add /path/to/learning-opportunities

The Codex marketplace includes three installable skills:

  • learning-opportunities: the core learning exercise skill
  • learning-opportunities-auto: optional post-commit prompting hook
  • orient: repo orientation generator

How It Works

  1. You complete significant architectural work (new files, schema changes, refactors, unfamiliar patterns)
  2. Claude asks if youd like a 10-15 minute learning exercise on a relevant topic
  3. You accept and Claude pauses, waiting for your input rather than providing the full answer
  4. You work through the exercise: predicting, generating, explaining, or tracing
  5. Claude reflects with you on what you observed or learned
  6. The skill suppresses further prompts if you decline or hit 2 exercises

The pause is intentional. The skill is designed to feel different from fast, fluent agentic coding. That friction is the point, as it creates space for active learning.

Customization

You can adapt the skill to your needs:

  • Include your technical expertise level and known languages to calibrate exercise difficulty
  • Ask Claude to save learning insights into your project CLAUDE.md
  • Adjust trigger conditions and exercise caps
  • Add project-specific examples or domain-specific retrieval questions
  • Add evaluation checks to measure how effectively the skill is working for you

Tradeoffs and Risks

  • Deliberate friction: The exercises intentionally slow you down. This is the feature, not a bug, but it may not fit every workflow
  • Claude Code and Codex only: No support for Gemini CLI, OpenCode, Copilot CLI, or Cursor currently
  • Optional by design: The skill only suggests exercises; it cant force learning engagement
  • Early-stage: Research-backed but relatively new in the AI coding tools space
  • Requires buy-in: Most effective if you commit to the 10-15 minute exercises consistently

Who Should Use It

  • Developers using Claude Code or Codex who want to build deeper understanding of the code they work with
  • Anyone concerned about skill erosion from over-reliance on AI code generation
  • Teams that want to make learning visible and valued in their engineering culture
  • Developers learning new languages, frameworks, or architectural patterns with AI assistance
  • Engineering leaders looking for a lightweight, research-backed intervention for team skill development

Why This Tool Rocks

  • Research-Backed: Every exercise type is grounded in peer-reviewed learning science, not guesswork
  • Context-Aware: Exercises draw from your actual project code, not generic examples
  • Non-Intrusive: Prompts only after natural breakpoints, respects decline decisions
  • Adaptive Difficulty: Can be calibrated to your existing expertise level
  • Team-Ready: Comes with validated measurement instruments and a team experiment playbook
  • Open Science: Created by a practicing psychological scientist who studies developer thriving in AI-assisted workflows
  • Companion Tools: orient and learning-goal extend the system for repo exploration and goal setting
  • Open Source: CC-BY-4.0 licensed, built on open science principles

🔗 GitHub: github.com/DrCatHicks/learning-opportunities

🔗 Author: drcathicks.com

🔗 Research: AI Skill Threat Open Access Measures

Crepi il lupo! 🐺