Skip to main content

The layero skill

A skill (Agent Skill) is a SKILL.md file with instructions that the agent reads when a task concerns Layero. There is one, it lives in the public repository LayeroInfra/layero-agents and it is the source of truth: the MCP server instructions, llms.txt, these pages and the block the CLI writes into AGENTS.md are checked against it by a gate. If something elsewhere says otherwise, the error is there, not in the skill.

What is inside​

skills/layero/
├── SKILL.md            ← the three paths (repository / folder / site), when to use
│                          which, sign-in and token, CLI JSON events, what not to do
└── references/
    ├── json-events.md  ← the schema of CLI events and error codes
    ├── layero-json.md  ← the layero.json file: framework, commands, output folder
    └── providers.md    ← GitHub, GitVerse, GitLab, GitFlic, SourceCraft

SKILL.md is short and answers "what to do"; the details live in references/, and the agent opens them only when needed.

How it is installed​

The method depends on the client. The commands come from the install table; here is what stands behind them.

Any agent that follows the .agents/skills standard:

npx skills add LayeroInfra/layero-agents

The command copies the skill into the project's .agents/skills/layero/. That is how Claude Code, Cursor, Codex and other clients that read this directory pick it up.

Claude Code — the plugin. The layero plugin includes both the skill and the MCP server connection:

claude plugin marketplace add LayeroInfra/layero-agents && claude plugin install layero@layero

Cursor — the layero-cursor plugin from the same repository: MCP, rules and the skill.

The plugins are generated from the skill by build-adapters.py in the layero-agents repository; they are never edited by hand.

npx layero@latest init and AGENTS.md​

npx layero@latest init

The command detects the framework, creates .layero/project.json and appends a short pointer to the project's AGENTS.md (or CLAUDE.md / .cursorrules if they exist): where the skill lives, how to install it, and the address https://layero.ru/llms.txt. The full instruction text is deliberately not there — one fact should have one home, and that is the skill. An agent that reads AGENTS.md goes to the skill for details, not to a stale copy.

If the skill is already installed in .agents/skills/, the pointer simply leads to it.

How to check the skill works​

Ask the agent: "how do I deploy this project to Layero?". The right answer starts with the question whether there is a connected repository and contains neither git init nor npm i -g. If the agent suggests creating a GitHub repository, the skill was not picked up.