What does it mean to apply TDD to writing skills?

It means you must first observe an agent fail a task (Red), write the documentation to address the failure (Green), and then refine the documentation for clarity and efficiency (Refactor).

How do I make my skills token-efficient?

Keep frequently-loaded skills under 200 words, move command details to tool help menus, and use explicit cross-references instead of force-loading files with @ syntax.

When should I create a new skill versus updating a project file?

Create a skill for reusable techniques and patterns that apply across projects; use project-specific files like CLAUDE.md for local conventions.

What is the recommended structure for a SKILL.md file?

A SKILL.md file should include YAML frontmatter (name and description), an Overview, 'When to Use' triggers, core patterns, and implementation examples.

Why shouldn't I summarize the workflow in the skill description?

If the description summarizes the workflow, Claude may follow the summary as a shortcut and ignore the more detailed instructions in the main body of the skill.

Writing Skills Framework

Name: Writing Skills Framework
Author: PREDICTif

byPREDICTif

•

学習とドキュメンテーション

Applies Test-Driven Development principles to create, optimize, and verify AI agent documentation and specialized skills.

The Writing Skills framework transforms documentation into a rigorous engineering process by applying the Red-Green-Refactor cycle to AI agent capabilities. Instead of writing passive guides, this skill directs users to create 'pressure scenarios' to observe agent failure, document specific solutions, and verify compliance. It provides a comprehensive standard for Claude Search Optimization (CSO), ensuring skills are discoverable, token-efficient, and structurally sound using standardized SKILL.md formats and Graphviz flowcharting.

主な機能

01Red-Green-Refactor workflow for process documentation and skill creation

02Claude Search Optimization (CSO) strategies for enhanced discovery

03Pressure-testing methodology to verify skill effectiveness before deployment

04Token-efficiency patterns to minimize context usage in long sessions

05Standardized SKILL.md structure with YAML frontmatter requirements

061 GitHub stars

ユースケース

01Developing new domain-specific capabilities for Claude Code or Codex agents

02Standardizing team-wide documentation practices for AI-assisted development

03Refactoring existing documentation to improve AI model adherence and performance

主な機能

01Red-Green-Refactor workflow for process documentation and skill creation

02Claude Search Optimization (CSO) strategies for enhanced discovery

03Pressure-testing methodology to verify skill effectiveness before deployment

04Token-efficiency patterns to minimize context usage in long sessions

05Standardized SKILL.md structure with YAML frontmatter requirements

061 GitHub stars

ユースケース

01Developing new domain-specific capabilities for Claude Code or Codex agents

02Standardizing team-wide documentation practices for AI-assisted development

03Refactoring existing documentation to improve AI model adherence and performance