skill-generator

Skill Generator

Generate production-quality Claude Code skills from any documentation source — repositories, websites, PDFs, or local files.

Overview

This plugin encodes the design patterns, quality standards, and generation workflow proven across 50+ skills in the Trail of Bits marketplace. It takes any documentation source and produces properly structured skills with:

• Sharp activation boundaries ("When to Use" / "When NOT to Use")
• Decision trees for branching logic
• Rationalizations to Reject (for security/audit skills)
• Cross-references between related skills
• Progressive disclosure (SKILL.md < 500 lines, details in references/)

Installation

# From the skills marketplace
claude skills install skill-generator

# Or manually add to .claude/settings.json
{
  "plugins": [
    "./plugins/skill-generator"
  ]
}

Usage

Generate Skills from Documentation

Generate skills from the Frida documentation at https://frida.re/docs/

Create skills from ./docs/ directory in this repo

Turn the OWASP Testing Guide into Claude Code skills

Create a Single Skill

Create a new skill for container security auditing

Improve an Existing Skill

Review this skill against quality standards and improve it

How It Works

1. Discovery

The generator analyzes your documentation source — scanning for topics, classifying them by type, and mapping relationships.

2. Planning

A generation plan is presented for your approval before any skills are created. You can add, remove, rename, or reclassify skills.

3. Two-Pass Generation

Pass 1: All skills are generated in parallel from templates, without cross-references. Pass 2: Cross-references are populated after all skills exist, ensuring no broken links.

4. Validation

Every generated skill is checked for structure, content quality, and link integrity.

Skill Types

Type	Template	Best For
Tool	templates/tool-skill.md	CLI tools, analysis software, frameworks
Technique	templates/technique-skill.md	Cross-cutting methodologies (harness writing, coverage)
Guidance	templates/guidance-skill.md	Behavioral rules, best practices, decision frameworks
Audit	templates/audit-skill.md	Security review, vulnerability detection, compliance

Structure

plugins/skill-generator/
├── .claude-plugin/
│   └── plugin.json
├── skills/
│   └── skill-generator/
│       ├── SKILL.md                        # Main entry point
│       ├── references/
│       │   ├── discovery.md                # Source analysis methodology
│       │   ├── quality-standards.md        # Quality patterns from 50+ skills
│       │   └── validation.md               # Validation checklist
│       └── templates/
│           ├── tool-skill.md               # CLI tools, software
│           ├── technique-skill.md          # Cross-cutting methods
│           ├── guidance-skill.md           # Behavioral guidance
│           └── audit-skill.md              # Security review
└── README.md

Quality Standards

Generated skills follow patterns proven across the Trail of Bits marketplace:

Pattern	Purpose
"When to Use" / "When NOT to Use"	Sharp activation boundaries
Rationalizations to Reject	Block shortcuts in security skills
Decision trees	Guide branching logic
Progressive disclosure	Core in SKILL.md, details in references/
Cross-references	Connect related skills
Evidence-based findings	Every claim needs proof (audit skills)
Anti-patterns with explanations	Teach WHY, not just WHAT

Author

Trail of Bits

License

See repository license.