The comprehensive guide to building Agent Skills for the agentskills.io open standard.
17 chapters. 82,000+ words. From your first skill to meta-skill routers, quality scoring systems, and multi-skill pipeline orchestration.
Skills work across Claude Code, Claude.ai, VS Code, Cursor, OpenAI Codex, Gemini CLI, and 20+ AI agent platforms.
yarmoluk.github.io/custom-skill-developer
| Section | Chapters | What You'll Build |
|---|---|---|
| Foundations | 1-3 | Your first working skill — from zero to production |
| Skill Anatomy | 4-7 | SKILL.md structure, YAML frontmatter, workflows, quality scoring rubrics |
| Advanced Patterns | 8-11 | Meta-skill routers, token efficiency, session logging, pipeline orchestration |
| Specialized Skills | 12-14 | Data transformation, code generation, and analysis/validation skills |
| Deployment | 15-17 | Installation, testing, debugging, publishing, and distribution |
Anthropic launched Agent Skills as an open standard. The format is simple — a folder with a SKILL.md file. But building production-quality skills that handle edge cases, score output quality, manage tokens efficiently, and chain together in pipelines requires patterns that aren't documented anywhere else.
This guide documents those patterns. Built from the experience of creating 20+ production Agent Skills for business operations.
- SKILL.md anatomy — Every section, why it exists, what breaks without it
- Quality scoring systems — 100-point rubrics with weighted categories and threshold gates
- Meta-skill routers — Consolidating 14+ sub-skills under one entry point with keyword-based routing
- Token efficiency — Progressive disclosure, lazy loading, skip-if-complete detection
- Pipeline orchestration — Chaining skills in dependency order with checkpoint patterns
- The 30-skill limit — Strategies for working within Claude Code's constraint
This guide is part of a three-layer stack for the Agent Skills ecosystem:
| Layer | Project | Purpose |
|---|---|---|
| Education | Custom Skill Developer | Teaches how to build production skills |
| Tooling | Skill Quality Analyzer | Audits skills against the spec (100-point rubric) |
| Product | Cognify Skills | 19 production business operations skills |
The guide ships with a full analytics layer built into the site — surfacing learning design quality alongside content depth.
| Metric | Value |
|---|---|
| Concepts | 200 concepts mapped across 12 categories with full dependency chains |
| Word Count | ~47,470 words across 17 chapters (chapter content) |
| Code Examples | 246 |
| Diagrams | 13 |
| Tables | 45 |
| Admonitions | 14 |
| Quiz Questions | 85 multiple-choice questions, 5 per chapter, Bloom's taxonomy aligned |
| Cognify Skills Audit | All 19 Cognify Skills scored against the 100-point rubric — 83.5 average |
The Learning Graph visualizes concept dependencies across all 12 categories, showing how foundational patterns build toward advanced pipeline orchestration. The Cognify Skills Scorecard publishes the full audit: per-skill scores, grade distribution, and rubric category breakdowns.
| Resource | Description |
|---|---|
| agentskills.io | The open standard specification |
| anthropics/skills | Anthropic's official example skills |
| Cognify Skills | 19 production business skills built to the standard |
| Skill Quality Analyzer | Audit skills with a 100-point scoring rubric |
Apache 2.0