Skill Bootstrap

Skill Bootstrap

Purpose

Generate new skills rapidly from natural language descriptions. This skill prioritizes speed and compliance with the official Claude Agent Skills specification over bulletproofing.

For critical discipline-enforcing skills, use skill-hardening after bootstrapping.

When to Use

• Creating utility skills quickly
• Bootstrapping API wrappers or reference guides
• Generating skill scaffolding from descriptions
• Rapid prototyping of new capabilities
• Team-specific or project-specific tools

When NOT to use:

• Discipline-enforcing skills (use skill-hardening instead)
• Complex pattern skills requiring rigorous testing
• Skills that agents must follow under pressure

Quick Start

User: "Create a skill for validating JSON schemas"

You: [Use this skill to generate SKILL.md]

Official Skill Format

File Structure

.claude/skills/my-skill/
├── SKILL.md              # Required, uppercase
├── examples/             # Optional
│   └── usage.md
├── templates/            # Optional
│   └── template.yaml
└── scripts/              # Optional
    └── helper.py

SKILL.md Frontmatter

---
name: skill-name-lowercase-hyphenated
description: What the skill does, when to use it, trigger keywords. Max 1024 characters.
allowed-tools: Read, Grep, Glob  # Optional: restrict tool usage
---

Required fields:

• name: Lowercase with hyphens
• description: 200-1024 characters, include what/when/triggers

Optional fields:

• allowed-tools: Comma-separated tool restrictions

Do NOT include:

• version, category, activation_criteria, triggers (as array)

Description Best Practices

Critical: The description determines when Claude activates the skill.

Good description:

description: Validate JSON documents against JSON Schema specifications. Use when working with JSON validation, API request/response validation, or data integrity checks. Supports draft-07, draft-2019-09 schemas. Triggers on "JSON schema", "validate JSON", "schema validation".

Bad description:

description: Helps with Excel files  # Too vague

Bootstrap Process

Step 1: Parse Request

Extract:

• Name: lowercase-hyphenated
• Purpose: core functionality
• Triggers: keywords and use cases
• Scope: included/excluded

Step 2: Generate SKILL.md

Structure:

---
name: extracted-name
description: [Comprehensive description]
---

# Skill Name

## Purpose
Brief explanation

## When to Use
- Use case 1
- Use case 2

## Core Instructions
Step-by-step guidance

### Patterns and Examples
Code examples

## Dependencies (Optional)
Required tools/libraries

## Version
v1.0.0

Step 3: Evaluate Supporting Files

Create only if they add significant value:

• examples/ → for detailed usage patterns
• templates/ → for reusable structures
• scripts/ → for helper utilities

Use progressive disclosure:

For examples, see [examples/usage.md](examples/usage.md)

Step 4: Validate

Check:

• ✓ SKILL.md exists (uppercase)
• ✓ Frontmatter has name and description
• ✓ Description is 200-1024 characters
• ✓ No invalid frontmatter fields
• ✓ Supporting files referenced correctly

Tool Restrictions (allowed-tools)

Use when you want to restrict Claude's tools:

---
name: safe-file-reader
description: Read files without making changes
allowed-tools: Read, Grep, Glob
---

When to use:

• Read-only skills
• Safety-critical workflows
• Limited-scope operations

After Bootstrapping

For Utility Skills

Deploy and use. Monitor for issues.

For Discipline-Enforcing Skills

Required: Use skill-hardening to bulletproof the skill with TDD.

⚠️ This skill enforces discipline. Run skill-hardening to test it:
- Create pressure scenarios
- Verify agents comply under stress
- Close loopholes

Example: Simple Skill

Request: "Create a skill for validating JSON schemas"

Generated SKILL.md:

---
name: json-schema-validator
description: Validate JSON documents against JSON Schema specifications. Use when working with JSON validation, API request/response validation, configuration file validation, or data integrity checks. Supports draft-07, draft-2019-09, and draft-2020-12 schemas. Triggers on "JSON schema", "validate JSON", "schema validation", "JSON structure check".
---

# JSON Schema Validator

## Purpose
Validate JSON documents against JSON Schema specifications.

## When to Use
- Validating API payloads
- Checking configuration files
- Data integrity verification

## Core Instructions

1. **Identify Schema**: Locate the JSON Schema
2. **Load JSON**: Read the document to validate
3. **Validate**: Compare against schema rules
4. **Report**: List validation errors with paths

### Basic Validation Pattern

```python
import jsonschema
import json

def validate_json(data, schema):
    try:
        jsonschema.validate(instance=data, schema=schema)
        return True, "Valid"
    except jsonschema.exceptions.ValidationError as e:
        return False, str(e)

Dependencies

• Python 3.7+
• jsonschema library

Version

v1.0.0

## Example: Multi-File Skill

**Request:** "Create a skill for API testing with examples"

**Generated structure:**

.claude/skills/api-testing/ ├── SKILL.md ├── examples/ │ └── rest-api-example.md └── templates/ └── test-template.yaml

## Best Practices

### Keep It Minimal
- Don't create files just because you can
- Each supporting file should add value
- Examples should teach, not overwhelm

### Description Writing
- Start with action verbs: "Analyze", "Generate", "Transform"
- Include specific file types: ".xlsx", ".json", ".py"
- List concrete use cases
- End with explicit triggers

### Progressive Disclosure
Supporting files load on-demand. Reference them:
```markdown
See [examples/usage.md](examples/usage.md) for detailed examples.

Integration with Other Skills

Version

v1.0.0 (2025-01-28) - Bootstrap-focused refactor of skill-creator