Skill Authoring Framework FAQs

Question 1

Why shouldn't I summarize the skill's process in the description?

Accepted Answer

Summarizing the workflow in the YAML description can cause Claude to take a shortcut and follow the brief summary instead of reading the full, detailed instructions in the skill body, leading to missed steps.

Question 2

What does it mean to apply TDD to writing skills?

Accepted Answer

It involves a Red-Green-Refactor cycle: creating a scenario where an agent fails to follow a process (Red), writing the skill documentation to guide them, verifying the agent now succeeds (Green), and then refining the text to close loopholes (Refactor).

Question 3

What is Claude Search Optimization (CSO)?

Accepted Answer

CSO is a set of practices including descriptive naming, keyword coverage, and specific 'Use when...' triggers that help Claude's internal search mechanism correctly identify and load the relevant skill for a task.

Question 4

How do I ensure my skills are token-efficient?

Accepted Answer

Efficiency is achieved by moving heavy reference material to separate files, using cross-references instead of repeating instructions, and keeping examples minimal—aiming for under 200 words for frequently-loaded skills.

Question 5

When should I create a new skill versus updating CLAUDE.md?

Accepted Answer

Create a skill for reusable techniques, patterns, or tools that apply broadly across projects. Use CLAUDE.md for project-specific conventions or one-off configurations that aren't useful in other contexts.

Skill Authoring Framework

Skill Authoring Framework

主な機能

ユースケース

主な機能

ユースケース