What is a SKILL.md file?

A SKILL.md file is the instruction manifest at the root of an AI agent skill folder. It starts with YAML frontmatter containing a name and a description, followed by markdown instructions the agent loads when the skill triggers. The format is an open standard supported by Claude Code, Codex CLI, Gemini CLI, GitHub Copilot, and Cursor.

How do I make a SKILL.md?

Write YAML frontmatter with a kebab-case name (max 64 characters) and a description (max 1024 characters) that says what the skill does and when to use it, then add imperative workflow instructions in the body. Keep the body under 500 lines and move detailed reference material into separate files under references/.

Should I put secrets in a public skill?

No. Public skills should describe credential handling and approval rules at a high level. They should never contain secret values, private keys, or instructions telling users to paste credentials into public files.

Skill Generator Knowledge in, working skill out

Turn anything you know into an AI agent skill.

Describe what the skill should do, optionally add links, documents, or notes, and get a complete skill folder: a spec-valid SKILL.md plus distilled reference files. One portable format that works in Claude Code, Codex CLI, Gemini CLI, GitHub Copilot, and Cursor.

Start Building How it works

One brief is all it takes

URLs / docs / paste optional knowledge

ZIP download drop into any agent

Spec-valid

Follows the Agent Skills standard

Valid frontmatter, a trigger-ready description, an imperative workflow body under 500 lines, and progressive disclosure into reference files.

Grounded

Built from your knowledge

Links, uploads, and pasted notes are distilled into clean reference files the agent reads on demand — no scraped page noise, no padding.

Tested

Trigger tests and a safety scan

Every bundle ships with should-trigger and should-not-trigger test prompts plus an automatic scan for risky patterns.

Step 1

What should this skill do?

One or two sentences is enough. Be specific about the job and the output you expect — the generator handles naming, structure, and formatting.

Step 2 · Optional

Add knowledge

Links, documents, or pasted notes get distilled into reference files inside the skill. The more real material you add, the more useful the skill.

Fetch a URL

Research a topic

Upload a document

TXT, MD, PDF, DOC, DOCX up to 8MB. Paste notes, SOPs, or docs

Attached

Knowledge sources

0 sources

No sources yet. You can generate from the brief alone, but sources make the skill far more useful.

Fine-tune (optional)

Skill name Lowercase letters, numbers, and hyphens. Becomes the folder name. Category (for the skills database)

When should it trigger — and when should it stay out of the way? Rules and constraints Allow executable script files when they genuinely help (advanced — most skills don't need them)

Your brief and sources are sent to the model, distilled into reference files, and written into a spec-valid SKILL.md you can download as a ZIP.

Generating

Building your skill

Preparing the brief and source material.

Brief Queued

Normalizing your brief and attached sources.

Distill Queued

Turning source material into clean reference files.

Write Queued

Writing SKILL.md with a trigger-ready description and workflow.

Done

Your skill is ready

Review the files, download the ZIP, or submit the skill to the database.

Bundle 0 files

SKILL.md plus references

Knowledge No reference files

Reference files appear here when sources are distilled into the skill.

Safety scan Not scanned

Every bundle is scanned for risky patterns automatically.

skill-name Claude Code · Codex · Gemini CLI · Copilot · Cursor

Generated skill

SKILL.md Basics

What makes a good agent skill?

A skill is a folder with a SKILL.md file at its root, plus optional references/ and scripts/. Agents only see the name and description until the skill triggers, so those two lines do most of the work — then the body guides the workflow, and reference files carry the deep knowledge.

A description that triggers

The description states what the skill does and exactly when to use it, with the keywords a user would actually type. Vague descriptions mean the skill never fires.

Progressive disclosure

The SKILL.md body stays lean — under 500 lines — and points the agent to reference files only when a task needs them. That keeps context cheap and skills fast.

One portable folder

The same skill folder works in Claude Code, Codex CLI, Gemini CLI, GitHub Copilot, and Cursor. No per-platform variants, no lock-in.

How It Works

From brief to bundle in three steps.

The generator distills your sources into clean reference files first, then writes a SKILL.md that routes the agent through them — the same structure hand-built skills from experienced authors use.

1. Describe the job

One clear paragraph about what the skill should do and what output you expect. Naming, structure, and metadata are handled for you.

2. Add knowledge (optional)

Fetch URLs, upload documents, run official-source research, or paste notes. Everything is distilled — scraped page noise and filler never reach the skill.

3. Review and download

Check the SKILL.md, the reference files, the trigger test prompts, and the safety scan. Download the ZIP or submit it to the skills database.