How to use the skill-development agent skill for Claude Code

What a skill is and what this one makes the agent do

A skill is a modular package that gives Claude specialised knowledge, workflows, and tools. Every skill has a required SKILL.md file that holds YAML frontmatter with a name and description, plus Markdown instructions. Skills can also bundle optional folders: scripts/ for executable code, references/ for documentation loaded as needed, and assets/ for files used in the output, such as templates.

Once this skill is active, Claude uses the skill-development guidance when you ask it to create or improve a skill. According to the guide, it asks about concrete example uses, works out which scripts, reference docs, or templates the skill needs, and then writes a SKILL.md with a clear name and trigger description. It also explains progressive disclosure: the name and description are always in context, the SKILL.md body loads when the skill triggers, and bundled resources load only as needed.

• Metadata (name and description) is always in context, at about 100 words.
• The SKILL.md body loads when the skill triggers and should stay under 5k words.
• Bundled resources load as needed; scripts can run without being read into context.

Install the skill-development skill

Download the ZIP from the skill's page on this directory. Then choose one of the following options.

For personal use across projects, unzip the download into ~/.claude/skills/. To limit it to one project, unzip it into that project's .claude/skills/ folder. If you use the Claude apps instead, upload the ZIP under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Trigger the skill with natural requests

You don't need a special command. The skill's description tells Claude when to use it, so ask in plain language. The description is written to match requests such as creating a skill, adding a skill to a plugin, writing a new skill, improving a skill description, organising skill content, or asking about skill structure and progressive disclosure.

Give Claude concrete context. The guide itself asks what the skill should do and what users would say to trigger it, so the more example requests you can supply, the better the result.

• "Create a skill for my plugin that formats changelog entries."
• "Improve the description of my skill so it triggers when users ask about invoice totals."
• "Help me organise the content of this skill so SKILL.md stays short."
• "Add a new skill to my plugin for generating release notes."

Work through the creation process with Claude

The guide sets out a creation process. Claude is expected to follow it in order and skip a step only when there is a clear reason. Step 1 is understanding the skill through concrete examples. Claude asks a few focused questions rather than many at once. Step 2 plans the reusable contents: for each example, it decides whether a script, reference file, or asset would save repeated work. Step 3 creates the directory structure. For Claude Code plugins, the guide shows a skills/ folder inside the plugin with one folder per skill.

The guide recommends keeping SKILL.md lean by moving detailed schemas, examples, and reference material into references/ files, and avoiding having the same information in both places. For large reference files, over 10k words, it suggests including grep search patterns in SKILL.md.

• Step 1: Understand the skill with concrete examples.
• Step 2: Plan scripts, references and assets for each example.
• Step 3: Create the skill structure.

Write a strong name and description

The name and description in the YAML frontmatter decide when Claude uses the skill, so they carry most of the weight. The guide asks for specific wording about what the skill does and when to use it, written in the third person, such as "This skill should be used when..." rather than "Use this skill when...".

• Be specific about the tasks and phrases that should trigger the skill.
• Use third-person wording such as "This skill should be used when...".
• Keep the main instructions short and move detail into references/.

Safety notes: scripts and network access

The skill-development skill itself does not run scripts and does not use the network, according to its listing. It produces skill files, so the usual review applies to what Claude writes. If the skill you build includes scripts, read them before you run them, since the guide notes that scripts may still need to be read by Claude for patching or environment-specific adjustments.

Check the skill's requirements before installing. This one needs Claude Code or a compatible SKILL.md-based agent runtime.

• Runs scripts: no
• Uses network: no
• Review any scripts in skills you generate before running them.