Skip to content

References

Curated references organized by chapter. Each entry includes a relevance note explaining why it matters for skill development.


Foundational Reading

These sources provide the intellectual and technical foundation for the Agent Skills ecosystem.

Dario Amodei

Machines of Loving Grace: How AI Could Transform the World for the Better Dario Amodei, October 2024

A 14,000-word vision of how powerful AI compresses decades of scientific, economic, and governance progress into years. The essay's core argument — that AI systems become exponentially more valuable when they can be specialized for domains while maintaining general reasoning — is the philosophical foundation for why skills exist. Skills are the mechanism by which a general-purpose model becomes a domain expert.

The Adolescence of Technology: Confronting and Overcoming the Risks of Powerful AI Dario Amodei, January 2026

The sequel to "Machines of Loving Grace" catalogs five civilization-level risks from powerful AI and proposes concrete countermeasures. Relevant to skill development because it frames the safety argument for structured, quality-gated agent behavior — skills constrain AI output to validated workflows rather than unbounded generation.


The Agent Skills Standard

Anthropic Official

Introducing Agent Skills Anthropic, October 2025 (updated December 2025)

The official product announcement for Agent Skills. Explains the product vision: skills as "onboarding guides" that transform a general-purpose agent into a domain specialist. Covers Claude.ai, Claude Code, the Agent SDK, and API integration.

Equipping Agents for the Real World with Agent Skills Barry Zhang, Keith Lazuka, Mahesh Murag — Anthropic Engineering, October 2025

The technical deep-dive into the skills architecture. Covers the SKILL.md format, progressive disclosure (the 3-tier token model), bundled scripts, security considerations, and the relationship between Skills and MCP. This is the most important technical reference for skill developers.

Agent Skills Open Standard Specification Anthropic, December 2025

The formal specification for the Agent Skills format. Defines frontmatter requirements (name, description, license, compatibility, metadata, allowed-tools), directory structure (scripts/, references/, assets/), progressive disclosure model, and validation rules. Cross-platform: Claude Code, Claude.ai, VS Code, Cursor, OpenAI Codex, Gemini CLI, and 20+ platforms.

Agent Skills Documentation Anthropic

Official documentation for using and creating skills. Covers installation, configuration, and authoring best practices.

Agent Skills Best Practices Anthropic

Anthropic's official authoring guidance for writing effective skills — conciseness, degrees of freedom, progressive disclosure, and common patterns.

Community Analysis

Agent Skills: Anthropic's Next Bid to Define AI Standards The New Stack

Strategic analysis of how Agent Skills parallels MCP as an industry standard play. Frames skills as the "what to do" complement to MCP's "what tools exist."

Agent Skills — Simon Willison Simon Willison, December 2025

Technical breakdown of the open standard release. Willison's analysis is consistently the most precise technical commentary in the AI ecosystem.


By Chapter

Chapters 1-3: Foundations

Reference Relevance
Agent Skills Specification The formal format definition — required reading for all skill authors
Anthropic Skills Repository Official example skills demonstrating patterns across creative, technical, and enterprise domains
What Are Skills? The conceptual overview — discovery, activation, execution model
Claude Code Documentation How Claude Code discovers and loads skills at session start

Chapters 4-7: Skill Anatomy

Reference Relevance
SKILL.md Format Specification Frontmatter field constraints, body content guidelines, validation rules
Skill Creator — Anthropic Anthropic's own meta-skill for creating skills — demonstrates anatomy best practices
Bloom's Taxonomy (2001 Revision) The six cognitive levels (Remember through Create) used in quality scoring rubric design
ISO 11179 Metadata Registry Standard The standard for precise, concise, distinct, non-circular definitions — used in glossary and rubric design

Chapters 8-11: Advanced Patterns

Reference Relevance
Progressive Disclosure in Skills The 3-tier token model: metadata (~100 tokens), instructions (<5,000 tokens), resources (as needed)
MCP — Model Context Protocol The complementary protocol to skills — MCP provides tool discovery, skills provide workflow knowledge
Dan McCreary — Intelligent Textbook Methodology The 12-step pipeline that demonstrated multi-skill orchestration at scale (19 skills, 200+ concepts per textbook)
Directed Acyclic Graphs — Wikipedia The data structure underlying learning graphs, dependency chains, and pipeline orchestration

Chapters 12-14: Specialized Skills

Reference Relevance
vis-network.js Documentation The library used for learning graph visualization in data format skills
Chart.js Documentation The charting library used in code generation skills for data visualization MicroSims
p5.js Reference The creative coding library used for interactive simulation MicroSims
JSON Schema Specification Used for metadata validation in analysis and code generation skills

Chapters 15-17: Deployment

Reference Relevance
Integrate Skills into Your Agent How to build a skills-compatible client — the integration specification
skills-ref Validation Library The official validation tool: skills-ref validate ./my-skill
Apache 2.0 License The recommended license for open-source skills (used by Anthropic's own skills)
Semantic Versioning 2.0.0 The versioning standard for published skill collections

Industry Context

Reference Relevance
Anthropic Opens Agent Skills Standard — Unite.AI Analysis of Anthropic's pattern of publishing open standards (MCP, then Skills) as industry infrastructure
Anthropic Launches Enterprise Agent Skills — VentureBeat Enterprise adoption angle and competitive positioning against OpenAI
GitHub Copilot Skills Support Microsoft/GitHub adoption of the Agent Skills standard in Copilot and VS Code
MkDocs Material Documentation The documentation framework used throughout this guide and the intelligent textbook methodology