How does it handle stale documentation?

It audits living docs against reality, updates diagrams and tables, marks completed plans as 'complete', and flags gaps where documentation is missing for existing code.

Does WF Gardener change my code's functionality?

No. Its mandate is specifically to clean and maintain without changing application behavior or making new design decisions. It focuses on legibility and documentation honesty.

How do I launch the gardener agent?

Once the skill is installed, you can launch it using the command: claude --agent wf:gardener.

What files does this skill primarily monitor?

It focuses on the design-docs/ directory, specifically ARCHITECTURE.md, PRINCIPLES.md, and implementation plans located at the repository and package levels.

What is considered 'LLM-generated technical debt'?

This includes defensive code for impossible cases, inconsistent naming conventions, and pattern drift that often accumulates during multiple automated coding sessions.

WF Gardener

Name: WF Gardener
Author: JSai23

byJSai23

•

学習とドキュメンテーション

Maintains codebase and documentation health by pruning technical debt and synchronizing system maps with current implementation.

WF Gardener is a specialized maintenance agent designed to eliminate the friction caused by LLM-generated technical debt and documentation drift. It continuously audits the design-docs directory, ensuring that ARCHITECTURE.md, PRINCIPLES.md, and living documents accurately reflect the current state of the code. By pruning stale plans, fixing documentation gaps, and cleaning up over-engineered code patterns, it ensures that subsequent AI development sessions inherit a clean, legible, and honest representation of the system, preventing the compounding errors that occur when docs and code diverge.

主な機能

01Documentation Health Auditing: Synchronizes architecture maps and design principles with the actual code state.

02Incremental Refinement: Executes small, test-backed changes to maintain system legibility without altering core application behavior.

03Technical Debt Cleanup: Identifies and removes LLM-generated defensive code, pattern drift, and stale comments.

04Drift Detection: Flags discrepancies between high-level implementation plans and actual logic to prevent misleading future sessions.

051 GitHub stars

06Plan Lifecycle Management: Automatically prunes completed plans and reconciles deviation logs to keep the project history clean.

ユースケース

01Context preparation before starting a new project phase by ensuring the project's design documentation is accurate for the next agent.

02Post-implementation cleanup after a major feature build to ensure living docs match the final delivery.

03Periodic repository maintenance to remove 'hallucinated' or over-engineered code snippets generated during rapid AI iterations.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jsai23/claude-tooling wf-gardener

For use in Claude.ai and ChatGPT

Download Skill