Agent Skills
Discoverable, reusable packages of instructions and resources that teach an agent a specialized workflow without loading the whole procedure into every session.
Use it when
- A workflow has a stable trigger, sequence, quality bar, and reusable supporting files.
- Teams repeatedly paste the same checklist or domain guidance into chats.
- The capability should be portable across projects or compatible agents.
Choose another layer when
- A short repository fact that every session needs.
- A deterministic lifecycle action that must run whether the model remembers it or not.
Production path
Build it in this order
- 01
Describe the trigger precisely
Give the Skill a clear name and a description that says both what it does and when it should run.
Quality check: Representative requests select the Skill; unrelated requests do not.
- 02
Use progressive disclosure
Keep the entrypoint focused, then link scripts, templates, references, and examples the agent should load only when relevant.
Quality check: The core workflow is understandable without injecting every reference into context.
- 03
Make success testable
Include validation steps, failure handling, and a concrete output contract. Test implicit and explicit invocation.
Quality check: The Skill produces the expected artifact on several real tasks and fails safely when prerequisites are missing.
Artifacts you may see
- SKILL.md
- scripts/*
- references/*
- assets/*
Connected layers
Evidence ledger
Primary documentation behind this guide
Provider-specific behavior is sourced to that provider. Open standards are sourced to their specifications and official maintainers.
Agent Skills specification
Agent Skills
The open SKILL.md package format and progressive-disclosure model.
Build skills for ChatGPT and Codex
OpenAI
Current OpenAI Skill structure, discovery, and authoring guidance.
Extend Claude with skills
Anthropic
Claude Code Skill discovery, configuration, supporting files, and invocation behavior.