A custom Claude code and Codex skill has been released, leveraging large language models to facilitate deliberate practice and adaptive feedback loops. By integrating Codex's generative capabilities with Claude's conversational interface, users can engage in targeted skill-building exercises, receiving real-time feedback and guidance to refine their knowledge and expertise.
Overview
This hybrid model holds promise for accelerating learning and skill acquisition. The skill uses an adaptive 'dynamic textbook' approach to help users integrate science-based expertise building exercises while doing agentic coding. When users complete architectural work, Claude offers optional 10-15 minute learning exercises grounded in evidence-based learning science.
What it does
The exercises use techniques like prediction, generation, retrieval practice, and spaced repetition to provide users with semi-worked examples from across their own project work. The skill pairs well with Learning-Goal, a skill that guides users through semi-structured, interactive learning goal-setting using the technique of Mental Contrasting with Implementation Intentions (MCII), an evidence-based exercise.
To install the Codex plugin, users can add the marketplace using the command codex plugin marketplace add https://github.com/DrCatHicks/learning-opportunities.git. For local development, users can add the plugin using codex plugin marketplace add /path/to/learning-opportunities. The Codex marketplace includes the core learning exercise skill, learning-opportunities-auto, and orient, a repo orientation generator.
Tradeoffs
The techniques in this skill are designed to counteract the risks of AI coding tools, such as the generation effect, fluency illusion, spacing effect, and metacognition. The skill interrupts the pattern of highly fluent and fast agentic coding, introducing a different 'mode' of interacting with Claude, which will intentionally feel different. This skill may be particularly useful for users who are experimenting with developing discrete projects with agentic coding that involve multiple unfamiliar languages, techniques, or architectural patterns.
In conclusion, the Claude Code and Codex skill offers a novel approach to skill development, leveraging large language models to facilitate deliberate practice and adaptive feedback loops. By integrating Codex's generative capabilities with Claude's conversational interface, users can engage in targeted skill-building exercises, receiving real-time feedback and guidance to refine their knowledge and expertise. This skill has the potential to accelerate learning and skill acquisition, and its techniques are designed to counteract the risks of AI coding tools.