Why does the skill sync to a cache?

Claude Code caches plugins separately from their source. The 'learn' skill uses rsync to push changes to the cache so your updates take effect immediately without needing to restart the environment.

Does it automatically commit changes to GitHub?

No. The skill modifies local files and displays a git diff, but it requires you to manually review and commit the changes to ensure accuracy and safety.

What does the 'learn' skill actually modify?

It modifies the Markdown definitions and instruction files for agents, skills, and commands located within your local fx-cc plugin marketplace directory.

When is the best time to use the /learn command?

Use it immediately after Claude makes a mistake, fails to use a tool correctly, or when you want to establish a permanent new rule for how an agent should handle specific tasks.

Adaptive Plugin Learner

Name: Adaptive Plugin Learner
Author: fx

byfx

•

생산성 및 워크플로

Updates and refines Claude Code plugin agents, skills, and commands based on real-time conversation feedback and corrections.

소개

The learn skill enables Claude to evolve by updating its own plugin definitions—including agents, skills, and commands—based on conversation history and direct user feedback. When an agent misbehaves or a skill fails to trigger correctly, this tool analyzes the root cause, modifies the source files in the fx-cc marketplace, and synchronizes those changes with the Claude Code cache for immediate application. It provides a structured workflow for the continuous improvement of AI workflows, ensuring that once a mistake is identified and corrected, the fix is persisted across future sessions while maintaining human oversight through mandatory manual Git reviews.

주요 기능

Automated plugin refinement based on natural language feedback
1 GitHub stars
Intelligent analysis of agent behavior and trigger conditions
Seamless synchronization between marketplace source and Claude Code cache
Precise file modification for agents, skills, and commands
Human-in-the-loop validation via mandatory manual Git review

사용 사례

Correcting an agent that consistently misses specific workflow steps during development
Updating skill descriptions to improve trigger reliability for specific CLI commands
Adding explicit behavioral prohibitions to prevent recurring AI errors or hallucinations

소개

주요 기능

Automated plugin refinement based on natural language feedback
1 GitHub stars
Intelligent analysis of agent behavior and trigger conditions
Seamless synchronization between marketplace source and Claude Code cache
Precise file modification for agents, skills, and commands
Human-in-the-loop validation via mandatory manual Git review

사용 사례

Correcting an agent that consistently misses specific workflow steps during development
Updating skill descriptions to improve trigger reliability for specific CLI commands
Adding explicit behavioral prohibitions to prevent recurring AI errors or hallucinations