|
| 1 | +<!-- |
| 2 | +SPDX-FileCopyrightText: Copyright © 2026 Caleb Cushing |
| 3 | +
|
| 4 | +SPDX-License-Identifier: CC0-1.0 |
| 5 | +--> |
| 6 | + |
| 7 | +# Skill Creator |
| 8 | + |
| 9 | +Guide for creating effective skills that extend Claude's capabilities. |
| 10 | + |
| 11 | +## About Skills |
| 12 | + |
| 13 | +Skills are modular, self-contained packages that extend Claude's capabilities by providing specialized knowledge, workflows, and tools. Think of them as "onboarding guides" for specific domains. |
| 14 | + |
| 15 | +### What Skills Provide |
| 16 | + |
| 17 | +1. **Specialized workflows** - Multi-step procedures for specific domains |
| 18 | +2. **Tool integrations** - Instructions for working with specific file formats or APIs |
| 19 | +3. **Domain expertise** - Company-specific knowledge, schemas, business logic |
| 20 | +4. **Bundled resources** - Scripts, references, and assets for complex tasks |
| 21 | + |
| 22 | +## Core Principles |
| 23 | + |
| 24 | +### Concise is Key |
| 25 | + |
| 26 | +The context window is a public good. Only add context Claude doesn't already have. |
| 27 | + |
| 28 | +**Default assumption: Claude is already very smart.** Challenge each piece of information: "Does Claude really need this explanation?" |
| 29 | + |
| 30 | +Prefer concise examples over verbose explanations. |
| 31 | + |
| 32 | +### Anatomy of a Skill |
| 33 | + |
| 34 | +``` |
| 35 | +skill-name/ |
| 36 | +├── SKILL.md (required) |
| 37 | +│ ├── YAML frontmatter (name, description) |
| 38 | +│ └── Markdown instructions |
| 39 | +└── Bundled Resources (optional) |
| 40 | + ├── scripts/ - Executable code |
| 41 | + ├── references/ - Documentation |
| 42 | + └── assets/ - Templates, images |
| 43 | +``` |
| 44 | + |
| 45 | +## SKILL.md Components |
| 46 | + |
| 47 | +### Frontmatter (YAML) |
| 48 | + |
| 49 | +```yaml |
| 50 | +--- |
| 51 | +name: skill-name |
| 52 | +description: What the skill does. Use when [activation trigger]. |
| 53 | +--- |
| 54 | +``` |
| 55 | + |
| 56 | +The description is the primary triggering mechanism. Include both what the skill does AND when to use it. |
| 57 | + |
| 58 | +### Body (Markdown) |
| 59 | + |
| 60 | +Instructions and guidance. Only loaded AFTER the skill triggers. |
| 61 | + |
| 62 | +## Bundled Resources |
| 63 | + |
| 64 | +### Scripts (`scripts/`) |
| 65 | + |
| 66 | +Executable code for tasks requiring deterministic reliability. |
| 67 | + |
| 68 | +### References (`references/`) |
| 69 | + |
| 70 | +Documentation loaded as needed into context. |
| 71 | + |
| 72 | +### Assets (`assets/`) |
| 73 | + |
| 74 | +Files used in output (templates, images, fonts). |
| 75 | + |
| 76 | +## Progressive Disclosure |
| 77 | + |
| 78 | +Skills use three-level loading: |
| 79 | + |
| 80 | +1. **Metadata** - Always in context (~100 words) |
| 81 | +2. **SKILL.md body** - When skill triggers (<5k words) |
| 82 | +3. **Bundled resources** - As needed |
| 83 | + |
| 84 | +Keep SKILL.md under 500 lines. Split content when approaching this limit. |
| 85 | + |
| 86 | +## Skill Creation Process |
| 87 | + |
| 88 | +1. **Understand** - Gather concrete usage examples |
| 89 | +2. **Plan** - Identify reusable scripts, references, assets |
| 90 | +3. **Initialize** - Create skill directory structure |
| 91 | +4. **Edit** - Implement resources and write SKILL.md |
| 92 | +5. **Package** - Bundle for distribution |
| 93 | +6. **Iterate** - Improve based on real usage |
| 94 | + |
| 95 | +## What NOT to Include |
| 96 | + |
| 97 | +- README.md |
| 98 | +- INSTALLATION_GUIDE.md |
| 99 | +- CHANGELOG.md |
| 100 | +- User-facing documentation |
| 101 | + |
| 102 | +Skills are for AI agents, not humans. Only include what Claude needs to do the job. |
0 commit comments