Skill Creator

Skill Creator

This skill provides guidance for creating effective skills following the Agent Skills specification.

About Skills

Skills are modular, self-contained packages that extend AI agent capabilities by providing specialized knowledge, workflows, and tools. Think of them as "onboarding guides" for specific domains or tasks—they transform a general-purpose agent into a specialized agent equipped with procedural knowledge that no model can fully possess.

What Skills Provide

1. Specialized workflows - Multi-step procedures for specific domains
2. Tool integrations - Instructions for working with specific file formats or APIs
3. Domain expertise - Company-specific knowledge, schemas, business logic
4. Bundled resources - Scripts, references, and assets for complex and repetitive tasks

Core Principles

Concise is Key

The context window is a public good. Skills share the context window with everything else the agent needs: system prompt, conversation history, other skills' metadata, and the actual user request.

Default assumption: The agent is already very smart. Only add context it doesn't already have. Challenge each piece of information: "Does the agent really need this explanation?" and "Does this paragraph justify its token cost?"

Prefer concise examples over verbose explanations.

Set Appropriate Degrees of Freedom

Match the level of specificity to the task's fragility and variability:

High freedom (text-based instructions): Use when multiple approaches are valid, decisions depend on context, or heuristics guide the approach.

Medium freedom (pseudocode or scripts with parameters): Use when a preferred pattern exists, some variation is acceptable, or configuration affects behavior.

Low freedom (specific scripts, few parameters): Use when operations are fragile and error-prone, consistency is critical, or a specific sequence must be followed.

Anatomy of a Skill

Every skill consists of a required SKILL.md file and optional bundled resources:

skill-name/
├── SKILL.md (required)
│   ├── YAML frontmatter metadata (required)
│   │   ├── name: (required)
│   │   ├── description: (required)
│   │   └── metadata.triggers: (optional, for auto-activation)
│   └── Markdown instructions (required)
└── Bundled Resources (optional)
    ├── scripts/          - Executable code (Python/Bash/etc.)
    ├── references/       - Documentation intended to be loaded into context as needed
    └── assets/           - Files used in output (templates, icons, fonts, etc.)

SKILL.md Frontmatter

Every SKILL.md must have YAML frontmatter with required and optional fields:

Trigger Configuration (metadata.triggers)

Skills can define auto-activation triggers in the metadata.triggers field:

metadata:
  triggers:
    type: domain           # "domain" (advisory) or "guardrail" (enforced)
    enforcement: suggest   # "suggest", "warn", or "block"
    priority: high         # "critical", "high", "medium", or "low"
    keywords:              # Exact substring matches (case-insensitive)
      - error
      - Result
      - error-stack
    intent-patterns:        # Regex patterns for intent detection
      - "\\b(handle|create)\\b.*?\\berror\\b"
      - "\\berror\\b.*?\\bhandling\\b"
    files:                  # Optional: file-based triggers
      include:
        - "**/src/**/*.rs"
      exclude:
        - "**/*.test.rs"
      content:
        - "use error_stack"

Trigger Types:

• keywords: Case-insensitive substring matching in user's prompt
• intent-patterns: Regex patterns to detect user intent (use \\b for word boundaries, .*? for non-greedy matching)
• files.include: Glob patterns for file paths
• files.exclude: Glob patterns to exclude (e.g., test files)
• files.content: Regex patterns to match file content

Enforcement Levels:

• suggest: Skill suggestion appears but doesn't block execution
• warn: Shows warning but allows proceeding
• block: Requires skill to be used before proceeding (guardrail)

Bundled Resources (optional)

Scripts (scripts/)

Executable code (Python/Bash/etc.) for tasks that require deterministic reliability or are repeatedly rewritten.

• When to include: When the same code is being rewritten repeatedly or deterministic reliability is needed
• Example: scripts/rotate_pdf.py for PDF rotation tasks
• Benefits: Token efficient, deterministic, may be executed without loading into context

References (references/)

Documentation and reference material intended to be loaded as needed into context.

• When to include: For documentation that the agent should reference while working
• Examples: references/finance.md for financial schemas, references/api_docs.md for API specifications
• Best practice: If files are large (>10k words), include grep search patterns in SKILL.md

Assets (assets/)

Files not intended to be loaded into context, but rather used within the output.

• When to include: When the skill needs files that will be used in the final output
• Examples: assets/logo.png for brand assets, assets/template.pptx for templates

Progressive Disclosure Design Principle

Skills use a three-level loading system to manage context efficiently:

1. Metadata (name + description) - Always in context (~100 words)
2. SKILL.md body - When skill triggers (<5k words recommended)
3. Bundled resources - As needed (unlimited)

Keep SKILL.md body under 500 lines. Split content into separate files when approaching this limit.

Pattern: High-level guide with references

# PDF Processing

## Quick start
Extract text with pdfplumber: [code example]

## Advanced features
- **Form filling**: See [FORMS.md](references/FORMS.md) for complete guide
- **API reference**: See [REFERENCE.md](references/REFERENCE.md) for all methods

Pattern: Domain-specific organization

bigquery-skill/
├── SKILL.md (overview and navigation)
└── references/
    ├── finance.md (revenue, billing metrics)
    ├── sales.md (opportunities, pipeline)
    └── product.md (API usage, features)

Skill Creation Process

Step 1: Understanding the Skill with Concrete Examples

To create an effective skill, clearly understand concrete examples of how the skill will be used:

• "What functionality should the skill support?"
• "Can you give some examples of how this skill would be used?"
• "What would a user say that should trigger this skill?"

Step 2: Planning the Reusable Skill Contents

Analyze each example to identify what scripts, references, and assets would be helpful:

• Scripts: Code that gets rewritten repeatedly (e.g., scripts/rotate_pdf.py)
• Assets: Boilerplate templates (e.g., assets/hello-world/ for frontend projects)
• References: Schemas and documentation (e.g., references/schema.md)

Step 3: Initializing the Skill

When creating a new skill from scratch, run the init command:

yarn agents:skill-management init <skill-name>

The command creates the skill directory in .claude/skills/ with a SKILL.md template and example resource directories.

Step 4: Edit the Skill

Write the Frontmatter

---
name: my-skill
description: What the skill does and when to use it. Include trigger keywords.
license: Apache-2.0
metadata:
  triggers:
    type: domain
    enforcement: suggest
    priority: medium
    keywords:
      - keyword1
      - keyword2
    intent-patterns:
      - "\\b(create|add)\\b.*?\\bsomething\\b"
---

Description best practices:

• Include both what the skill does and specific triggers/contexts
• Include all "when to use" information here - the body is only loaded after triggering
• Max 1024 characters

Write the Body

Write instructions for using the skill. Keep under 500 lines.

Step 5: Generate and Validate Skill Rules

After creating/modifying skills, validate and regenerate the skill-rules.json:

yarn agents:skill-management validate
yarn agents:skill-management generate-skill-rules

Step 6: Test the Skill

Test with a specific prompt:

echo '{"session_id":"test","prompt":"your test prompt","cwd":"."}' | \
  yarn workspace @local/claude-hooks run:skill

Debug matching logic:

echo '{"session_id":"test","prompt":"your test prompt","cwd":"."}' | \
  yarn workspace @local/claude-hooks dev:skill

Reference Files

For detailed information on specific topics, see:

• references/workflows.md: Sequential and conditional workflow patterns
• references/output-patterns.md: Template and example patterns for consistent output
• references/patterns-library.md: Ready-to-use regex and glob patterns for triggers
• references/troubleshooting.md: Debugging guide for skill activation issues

Testing Checklist

• Skill file created in .claude/skills/{name}/SKILL.md
• Proper frontmatter with name and description
• Triggers configured in metadata.triggers
• skill-rules.json regenerated
• Keywords tested with real prompts
• Intent patterns tested with variations
• SKILL.md under 500 lines
• Reference files created if needed