Agent Skills: Portable Contract and Runtime Boundary
A skill is not a long prompt with a better filename. It is a discoverable package of instructions, resources, and executable helpers that enters an agent's context through a runtime contract. Define an agent skill without confusing it with a prompt, repository instructions, a tool, a hook, a subagent, or a plugin. Read the portable SKILL.md contract and separate it from runtime-specific extensions. Explain discovery, selection, activation, resource loading, tool use, and verification as distinct lifecycle stages. Validate a skill package before a runtime places it in an agent's catalog. Choose between a skill, MCP tool, hook, subagent, or ordinary code for a concrete task. Do this before the long explanation. You will create a small skill, install the complete reviewer bundle into a real agent host, invoke it, verify the result, and remove it. This proves the lifecycle with an observable result. The real-host checkpoint requires Node.js, npx, Python 3, one selected skill-capable host, and write access to the project or user scope you choose in the installer. Verify the local commands first: Decide which host and scope you will use before installation. If any requirement is unavailable, read this lesson on the website or continue with the manual package exercise below. That fallback teaches the contract, but it does not prove host discovery, invocation, bundled-script execution, or uninstall…
Agent Skills: Portable Contract and Runtime Boundary: A skill is not a long prompt with a better filename. It is a discoverable package of instructions,…
This free lesson is part of the AI Engineering from Scratch curriculum. Read the full explanation, run the lesson code, and verify the result in the interactive reader or from the repository source.
Browse the complete course catalog or open this lesson on GitHub.