# Agents

> Install the NatUI Agent Skill and connect coding agents to current machine-readable documentation.

Canonical: https://natui.dev/docs/agents

NatUI provides an installable Agent Skill and a complete Markdown view of this
documentation. Use the skill for reusable implementation guidance, then use
`llms.txt` to find the current pages for the task. They complement each other:
the skill teaches the workflow, while the documentation supplies the current
API and platform details. The skill follows the
[Agent Skills specification](https://agentskills.io/specification). The
repository's `/skills` directory is a distribution catalog, so install the
skill into the client that will use it.

## Install the NatUI skill [#install-the-natui-skill]

Install the skill from GitHub with the
[open `skills` CLI](https://github.com/vercel-labs/skills):

```bash
npx skills add https://github.com/floklein/natui --skill natui
```

The interactive flow detects compatible agents and lets you choose project or
global scope. Review the
[skill source](https://github.com/floklein/natui/tree/main/skills/natui)
before installing it. The NatUI skill contains instructions and references,
with no executable scripts.

## Give agents current documentation [#give-agents-current-documentation]

| Resource                                            | Use                                                                                                        |
| --------------------------------------------------- | ---------------------------------------------------------------------------------------------------------- |
| [`https://natui.dev/llms.txt`](/llms.txt)           | Start here. It is the compact index of documentation pages and descriptions.                               |
| [`https://natui.dev/llms-full.txt`](/llms-full.txt) | Load the complete documentation corpus when a task genuinely spans most of NatUI.                          |
| `https://natui.dev/docs/<page>.md`                  | Fetch one clean Markdown page, such as [`components/inputs`](https://natui.dev/docs/components/inputs.md). |
| `Accept: text/markdown`                             | Request Markdown from the normal documentation URL when the client supports content negotiation.           |

Resolve relative links from `llms.txt` against `https://natui.dev`. Prefer the
compact index plus a few targeted pages over loading the full corpus into every
task. The compact index follows the
[`llms.txt` proposal](https://llmstxt.org/). `llms-full.txt` is a convenient
complete export, not a required part of that proposal.

```bash
curl -H "Accept: text/markdown" \
  https://natui.dev/docs/guides/controlled-state
```

Every documentation page also exposes **Copy Markdown** and **View Markdown**
actions for interactive use.

## Agent workflow [#agent-workflow]

1. Install or invoke the `natui` skill.
2. Read [`llms.txt`](/llms.txt) and select only the relevant setup, component,
   guide, and status pages.
3. Inspect the current checkout before editing. NatUI is in alpha and is not a
   published registry package.
4. Implement with React state and NatUI components. Do not introduce DOM
   elements, CSS, or browser event APIs.
5. Build the matching native host and validate on the target platform.
6. Use a real native window before claiming visible or interactive behavior.

Use this bootstrap prompt after installation:

```text
Use the installed NatUI skill to implement this task. Begin with
https://natui.dev/llms.txt, fetch only the relevant Markdown pages, inspect the
current checkout because NatUI is in alpha, and verify the result on the
requested native platform.
```

## High-value entry points [#high-value-entry-points]

* [Source setup](/docs/start/source-setup)
* [First app](/docs/start/first-app)
* [Component catalog](/docs/components)
* [Controlled state](/docs/guides/controlled-state)
* [Testing and debugging](/docs/guides/testing-and-debugging)
* [Platform support](/docs/status/platform-support)
* [Verification status](/docs/status/verification)