Skill Builder

Create Claude Code skills that extend capabilities with specialized knowledge, workflows, and tools.

Quick Commands

Core Principles

1. Concise is key - Claude is smart; only add what it doesn't know
2. Progressive disclosure - Metadata always loaded, SKILL.md on trigger, references on-demand
3. Appropriate freedom - High for flexible tasks, low for fragile operations

Skill Creation Process

Step 1: Clarify Requirements

Ask the user:

• What task should the skill automate?
• What triggers should invoke it? (keywords, file types, scenarios)
• What outputs are expected?
• Can you show example use cases?

Step 2: Plan Resources

Analyze each use case to identify:

• scripts/ - Repetitive code that should be deterministic
• references/ - Documentation too long for SKILL.md
• assets/ - Templates, images, boilerplate for output

Step 3: Initialize

python scripts/init_skill.py my-skill --path ~/.claude/skills

Step 4: Edit SKILL.md

Frontmatter (Critical)

---
name: my-skill
description: What this skill does and WHEN to use it. Include triggers, file types, scenarios. This is the primary trigger mechanism. Max 1024 chars.
---

Description rules:

• Include BOTH what the skill does AND when to use it
• List specific triggers (file types, keywords, scenarios)
• No < or > characters
• Max 1024 characters

Body Patterns

Pattern 1: Workflow-Based (sequential processes)

## Workflow
1. Analyze input -> Run `scripts/analyze.py`
2. Transform -> Apply rules
3. Output -> Save result

Pattern 2: Task-Based (tool collections)

## Tasks
### Create New
[instructions]
### Edit Existing
[instructions]

Pattern 3: Reference-Based (standards/guidelines)

## Guidelines
### Colors
### Typography

Step 5: Add Resources

Scripts (scripts/)

For deterministic, repetitive operations.

• Test scripts before packaging
• Make executable: chmod +x script.py

References (references/)

Documentation loaded on-demand.

• For content >100 lines, add table of contents
• Split by domain/variant when >500 lines
• Avoid duplication with SKILL.md

Assets (assets/)

Files used in output (not loaded into context).

• Templates, images, fonts
• Boilerplate code

Step 6: Validate & Package

python scripts/quick_validate.py my-skill/
python scripts/package_skill.py my-skill/ ./dist

Frontmatter Rules

Progressive Disclosure Patterns

Keep SKILL.md under 500 lines. Split to references/ when needed.

High-level guide with references:

## Quick Start
[Essential workflow]

## Advanced
- **Forms**: See `references/forms.md`
- **API**: See `references/api.md`

Domain-specific organization:

skill/
├── SKILL.md (overview + navigation)
└── references/
    ├── aws.md
    ├── gcp.md
    └── azure.md

What NOT to Include

• README.md (redundant with SKILL.md)
• CHANGELOG.md
• INSTALLATION_GUIDE.md
• User-facing documentation

Skills are for AI agents, not humans.

References

• references/skill-creation-guide.md - Complete skill creation guide
• references/output-patterns.md - Template and example patterns
• references/workflows.md - Sequential and conditional workflow patterns