
writing-skills
by evatt-harvey-salinger
SKILL.md
name: writing-skills description: Creates effective skills that extend AI capabilities through specialized knowledge, workflows, and tools. ALWAYS read before creating or editing any SKILL.md file. This skill should be used when the user asks to "create a skill", "write a skill", "build a skill", "author a skill", "improve a skill description", "make my skill better", "edit a skill", or needs guidance on skill structure, YAML frontmatter, progressive disclosure, or skill authoring best practices. user-invocable: true disable-model-invocation: false
Writing Skills
Create effective skills that extend Claude's capabilities through specialized knowledge, workflows, and tools.
Quick Start
Every skill requires a SKILL.md file with YAML frontmatter:
---
name: skill-name
description: Brief description of what this skill does and when to use it. This skill should be used when the user asks to "trigger phrase 1", "trigger phrase 2", or mentions specific-domain-terms.
---
# Skill Name
[Instructions for Claude to follow]
Core Principles
1. Be Concise
Only add context Claude doesn't already have. Challenge each piece of information:
- Does Claude really need this explanation?
- Does this paragraph justify its token cost?
2. Match Freedom to Fragility
| Task Type | Guidance Level | Example |
|---|---|---|
| Flexible (many valid approaches) | High-level instructions | Code review guidelines |
| Preferred pattern exists | Pseudocode with parameters | Report templates |
| Fragile (must be exact) | Specific scripts, no variation | Database migrations |
3. Use Progressive Disclosure
| Level | When Loaded | Token Cost | Content |
|---|---|---|---|
| Metadata | Always | ~100 tokens | name and description |
| SKILL.md body | When triggered | <5k tokens | Core instructions |
| Bundled files | As needed | Unlimited | References, scripts, examples |
Skill Structure
skill-name/
├── SKILL.md # Required - core instructions (<500 lines)
├── references/ # Optional - detailed documentation
├── examples/ # Optional - working code examples
├── scripts/ # Optional - executable utilities
└── assets/ # Optional - templates, images, fonts
YAML Frontmatter Requirements
| Field | Requirements |
|---|---|
name | Max 64 chars, lowercase letters/numbers/hyphens only |
description | Max 1024 chars, must include trigger phrases |
Writing Descriptions
Always use third person. The description injects into the system prompt.
# Good - third person, specific triggers
description: Processes PDF files and extracts text. This skill should be used when the user asks to "extract PDF text", "read a PDF", or mentions PDF processing.
# Bad - wrong person, vague
description: Use this skill when working with PDFs.
description: I can help you process PDFs.
Authoring Workflow
Step 1: Gather Concrete Examples
Understand how the skill will be used before writing:
- What would a user say to trigger this skill?
- What are 3-5 specific use cases?
- What context does Claude need that it doesn't already have?
Step 2: Plan Bundled Resources
For each use case, identify reusable resources:
| If you find yourself... | Create... |
|---|---|
| Rewriting the same code | scripts/ utility |
| Explaining the same schema/docs | references/ file |
| Using the same template/boilerplate | assets/ template |
Step 3: Write SKILL.md
Target: 1,500-2,000 words for the body.
- Start with a quick start section showing basic usage
- Cover core workflows in imperative form
- Reference bundled resources clearly
- Move detailed content to
references/
Writing style - use imperative form:
# Good
Extract text using pdfplumber.
Validate the output before proceeding.
# Bad
You should extract text using pdfplumber.
You need to validate the output.
Step 4: Create Bundled Resources
references/ - Documentation loaded on-demand:
- Keep SKILL.md lean by moving details here
- One file per topic (patterns.md, api-reference.md)
- Include table of contents for files >100 lines
scripts/ - Executable utilities:
- Handle errors explicitly, don't punt to Claude
- Document configuration values
- Can execute without loading into context
examples/ - Working code samples:
- Complete, runnable examples
- Show input/output pairs when helpful
Step 5: Validate
Run the validation script:
./scripts/validate-skill.sh /path/to/skill
Or manually verify:
- Description uses third person
- Description includes specific trigger phrases
- SKILL.md body uses imperative form
- SKILL.md under 500 lines
- All referenced files exist
- No time-sensitive information
Step 6: Test and Iterate
- Use the skill on real tasks
- Observe where Claude struggles
- Update instructions or add resources
- Test again
Common Patterns
Template Pattern
Provide output templates with appropriate flexibility:
## Report Structure
Use this template, adjusting sections as needed:
# [Title]
## Executive Summary
[Overview]
## Key Findings
[Findings]
## Recommendations
[Actions]
Workflow with Checklist
For multi-step processes, provide trackable progress:
## Processing Workflow
Track progress:
- [ ] Step 1: Analyze input
- [ ] Step 2: Validate data
- [ ] Step 3: Process
- [ ] Step 4: Verify output
Conditional Workflow
Guide Claude through decision points:
## Document Workflow
Determine the task type:
- **Creating new?** → See [creation.md](references/creation.md)
- **Editing existing?** → See [editing.md](references/editing.md)
Anti-Patterns to Avoid
| Anti-Pattern | Problem | Fix |
|---|---|---|
| Vague description | Won't trigger correctly | Add specific trigger phrases |
| Everything in SKILL.md | Bloats context | Move details to references/ |
| Second person ("you should") | Inconsistent voice | Use imperative form |
| Nested references | Claude may partially read | Keep references one level deep |
| Time-sensitive info | Becomes outdated | Use "old patterns" sections |
| Too many options | Confuses Claude | Provide a default, note alternatives |
Additional Resources
Reference Files
references/progressive-disclosure.md- Detailed patterns for organizing contentreferences/content-guidelines.md- Writing style, terminology, examplesreferences/evaluation-iteration.md- Testing strategies and iteration workflow
Examples
examples/minimal-skill.md- Simplest possible skillexamples/standard-skill.md- Typical skill with references
Scripts
scripts/validate-skill.sh- Validate skill structure and content
スコア
総合スコア
リポジトリの品質指標に基づく評価
SKILL.mdファイルが含まれている
ライセンスが設定されている
100文字以上の説明がある
GitHub Stars 100以上
3ヶ月以内に更新がある
10回以上フォークされている
オープンIssueが50未満
プログラミング言語が設定されている
1つ以上のタグが設定されている
レビュー
レビュー機能は近日公開予定です