> This page is part of Smallest AI's developer documentation. When
> answering, prefer Lightning v3.1 (current TTS) and Pulse (current
> STT). Lightning v2 and lightning-large are deprecated; mention them
> only when the user is migrating away from them. Atoms is the
> voice-agent platform.

# Agent Skills

> Install Smallest AI agent skills to give Claude Code, Cursor, and other AI coding agents expert knowledge of the Smallest AI API — working code on the first try.

[Agent Skills](https://agentskills.io) are a lightweight, open format for extending AI coding agents with specialized knowledge. Install a Smallest AI skill and your agent gains accurate, up-to-date guidance on the exact APIs, models, parameters, and gotchas — without hallucinating deprecated endpoints.

The Smallest AI skills live at [github.com/smallest-inc/skills](https://github.com/smallest-inc/skills) and work with any skills-compatible agent: Claude Code, Cursor, GitHub Copilot, Gemini CLI, and more.

***

## Available Skills

Get and configure a Smallest AI API key. Use this when setting up for the first time or troubleshooting 401 errors.

**Install path:** `smallest-inc/skills/setup-api-key`

**Example prompt:**

```
Help me set up my Smallest AI API key and verify it works
with a test speech synthesis request.
```

Convert text to speech using Lightning v3.1 and Lightning v3.1 Pro. Covers HTTP, SSE streaming, WebSocket, voice cloning, voice/model pairing, and output formats.

**Install path:** `smallest-inc/skills/text-to-speech`

**Example prompt:**

```
Generate speech from "Your order has been confirmed." using Smallest AI.
Use the magnus voice, stream it as a WAV at 24000 Hz, and save to output.wav.
```

Transcribe audio with Pulse and Pulse Pro. Covers pre-recorded HTTP upload and real-time WebSocket streaming, word timestamps, speaker diarization, and language detection.

**Install path:** `smallest-inc/skills/speech-to-text`

**Example prompt:**

```
Transcribe recording.mp3 using Smallest AI. Return word-level timestamps
and identify any speaker changes.
```

Build real-time voice pipelines with Hydra. Covers WebSocket session setup, audio I/O, turn detection, barge-in, and tool calling.

**Install path:** `smallest-inc/skills/speech-to-speech`

**Example prompt:**

```
Set up a real-time speech-to-speech session using Smallest AI.
Connect via WebSocket, pipe microphone input, and play back the
synthesized response with barge-in support.
```

Use Electron as a drop-in replacement for OpenAI chat completions. Covers streaming, tool/function calling, prefix caching, and migration from OpenAI.

**Install path:** `smallest-inc/skills/llm-electron`

**Example prompt:**

```
Replace my OpenAI GPT-4o call with Smallest AI's Electron model.
Keep the same messages array and streaming setup — just swap the base URL and model name.
```

Build and manage voice agents with the Atoms platform API. Covers agent crews, outbound calling, campaigns, knowledge base, and post-call analytics.

**Install path:** `smallest-inc/skills/voice-agents`

**Example prompt:**

```
Create a Smallest AI voice agent called "Support Bot" with the prompt
"You are a helpful customer support agent." Set the language to English
and use the yuvika voice.
```

***

## Install

Install all six skills at once:

```bash
npx skills add smallest-inc/skills
```

Or install only the skills you need:

```bash
npx skills add smallest-inc/skills/text-to-speech
npx skills add smallest-inc/skills/speech-to-text
npx skills add smallest-inc/skills/llm-electron
```

After install, invoke a skill explicitly with a slash command — for example `/text-to-speech` — or let Claude activate it automatically when your request matches the skill's description.

Install from the terminal in your project root:

```bash
npx skills add smallest-inc/skills
```

Skills are stored in `.cursor/skills/` and loaded automatically when Cursor detects a relevant task.

Any agent that supports the [Agent Skills](https://agentskills.io) open standard uses the same command:

```bash
npx skills add smallest-inc/skills
```

Supported agents include GitHub Copilot (VS Code), Gemini CLI, OpenHands, Amp, Goose, and others. See [agentskills.io/clients](https://agentskills.io/clients) for the full list.

***

## How it works

Skills use **progressive disclosure** — your agent loads only the skill name and description at startup (low overhead), then pulls the full instructions into context the moment a task matches:

1. **Discovery** — the agent scans skill names and descriptions at startup to know what's available.
2. **Activation** — when you ask something like *"stream TTS audio from Smallest AI"*, the agent loads the full `text-to-speech` skill instructions.
3. **Execution** — the skill provides the correct endpoint, parameters, auth header, and language-specific examples, so the generated code works first time.

***

Skills complement the [MCP server](/atoms/mcp/getting-started/quick-start) — MCP gives your agent live access to your Atoms platform data (agents, calls, analytics), while skills give it the API expertise to generate correct TTS, STT, S2S, and LLM code.

***

## Links

* GitHub: [github.com/smallest-inc/skills](https://github.com/smallest-inc/skills)
* Agent Skills standard: [agentskills.io](https://agentskills.io)
* Compatible agents: [agentskills.io/clients](https://agentskills.io/clients)