Skill
also called Capability, Playbook
A packaged procedure the agent can load when the task calls for it.
A reusable bundle of instructions, examples and sometimes scripts that teaches an agent how to do a specific job — a deployment checklist, a code-review rubric, a data-cleaning routine. Loaded on demand rather than always present.
Skills are how you get consistency across a team without stuffing every process into the system prompt. The agent sees a short description and pulls in the full procedure only when relevant.
The design principle is progressive disclosure applied to procedures. A skill keeps the main context small — the agent sees that a skill exists and what it is for — and pays the token cost of the full procedure only when the task actually calls for it.
A release procedure nobody can skip steps in
Preparing a release involves eight steps that people forget in the same order every time.
Write it as a skill: the commands, the checks, the rollback. Now the agent runs the procedure and the procedure is reviewable in a diff. Encoding the checklist once beats hoping everyone remembers it.
- blogEquipping agents for the real world with Agent Skills ↗
Defines Skills as discoverable folders of instructions, scripts and resources loaded progressively — the modular-reusable-procedure half of agent scaffolding.
- docsClaude Code overview ↗
A first-party description of the surfaces an agentic coding tool ships on, including parallel sessions and long-running tasks. Not evidence that it improves productivity.
AGENTS.md · Project instructions file
A plain file in the repo that tells agents how this project works.
Progressive Disclosure
Reveal detail only when it becomes relevant.
Context Pointer · Reference
A cheap reference to something the agent can fetch if needed.
Harness · Scaffold
Everything around the model that makes it an agent.