AI Dimensie
Wave 11·Docs · part of AI Dimensie·Why Docs? →·⌘K jump

Build your own aidimension skill and publish it to the registry.

Author a Skill

A skill is a Markdown file with frontmatter that teaches an AI agent how to use a tool, library, or API. aidimension skills follow a strict spec so they work in every agent.

File format#

Every skill is a single .md file with:

---
name: my-skill
version: 0.1.0
description: One-line description (max 80 chars)
author: Your Name
agent: [claude, cursor, gpt, windsurf, aider]
license: MIT
---

# My Skill

[Documentation here]

Required frontmatter#

FieldDescription
nameKebab-case, lowercase, no spaces. Max 64 chars.
versionSemver (e.g. 0.1.0)
descriptionOne sentence. Max 80 chars. Shown in search results.
authorYour name or org
agentArray of supported agents
licenseMIT, Apache-2.0, etc.

Optional frontmatter#

FieldDescription
homepageURL to your project
repositoryURL to source
keywordsArray of search keywords
triggersPhrases that auto-activate the skill
dependenciesOther skills this one needs

Required sections#

The Markdown body must have these sections in this order:

1. Title and intro#

# My Skill

You are an expert at [task]. Use this skill when [trigger condition].

2. When to use this skill#

## When to use this skill

Activate this skill when the user wants to:
- [Use case 1]
- [Use case 2]
- [Use case 3]

3. Canonical patterns#

Show 2-3 working examples. These are the most important part — agents pattern-match on them:

## Canonical patterns

### Simple case
\`\`\`tsx
[minimal example]
\`\`\`

### Common case
\`\`\`tsx
[realistic example]
\`\`\`

### Edge case
\`\`\`tsx
[tricky case]
\`\`\`

4. API reference#

## API reference

### functionName(args)
- `arg1` (type, required) — description
- `arg2` (type, optional) — description
- Returns: type — description

List things agents should NOT do:

## Anti-patterns

- ❌ Don't use [wrong way] — use [right way] instead
- ❌ Don't forget to [step]

Style guide#

  • Use ## for sections, not # (the title is the only #)
  • Keep code examples under 20 lines each
  • Use TypeScript by default (works for any agent)
  • Don't use emojis in headings — they slow down parsing
  • One blank line between sections
  • Use tables for reference material
  • Use bullet lists for "what to do" / "what not to do"

Length#

Aim for 3-8 KB of Markdown. Less than 2 KB and you don't have enough context; more than 16 KB and the agent will get distracted.

Testing locally#

Before publishing, test the skill in your own agent:

  1. Save the file as my-skill.md
  2. Install it (see Install a Skill)
  3. Try 5-10 different prompts that should trigger it
  4. Try 5-10 prompts that should NOT trigger it
  5. Iterate

Publishing#

When the skill is ready:

  1. Open a PR at github.com/pathtoresiliencebv/aidimension-skills with your .md file in skills/
  2. Add an entry to skills/registry.json:
{
  "name": "my-skill",
  "version": "0.1.0",
  "description": "...",
  "author": "you",
  "repository": "https://github.com/you/my-skill",
  "license": "MIT",
  "path": "skills/my-skill.md"
}
  1. The aidimension team reviews within 3 business days
  2. Once merged, it's available at https://skills.aidimension.io/my-skill.md and via npx @aidimension/cli@latest skills install my-skill

Versioning#

Bump the version when you update the skill:

  • 0.0.x — typos, small clarifications
  • 0.x.0 — new examples, sections
  • x.0.0 — breaking changes (different examples, new API)

License#

The official registry accepts MIT, Apache-2.0, and BSD-3-Clause. Other licenses require manual review.

Examples#

The best way to learn is by reading existing skills:

Next steps#