Steer logo

Steer

Community
disler
steer

macOS GUI automation CLI. Use steer to see the screen, click elements, type text, send hotkeys, scroll, drag, manage windows and apps, run OCR on Electron apps, and wait for UI conditions.

Overview

Publisherdisler
Repositorymac-mini-agent
Skill namesteer
Stars
274
Forks
74
Bundled files
Instructions only
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by disler on GitHub. Read the source before you install it.

Installation

Install the Steer AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/disler/mac-mini-agent/tree/main/.claude/skills/steer
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/disler/mac-mini-agent.git /tmp/mac-mini-agent
mkdir -p .claude/skills
cp -r /tmp/mac-mini-agent/.claude/skills/steer .claude/skills/steer
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Steer in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Steer on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Steer is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Steer — macOS GUI Automation

Binary: apps/steer/.build/release/steer

Run steer --help and steer help <command> to learn each command's flags before using it.

Commands

CommandPurpose
seeTakes a screenshot (PNG) and walks the accessibility tree. Screenshot always succeeds. Elements are best-effort — may be empty for Electron apps. Pass --ocr to fall back to OCR when the tree is empty.
clickClick by element ID, label, or coordinates
typeType text into focused element or a target
hotkeyKeyboard shortcuts (cmd+s, return, escape, etc.)
scrollScroll up/down/left/right
dragDrag between elements or coordinates
appsList, launch, or activate apps
screensList displays with resolution and scale
windowMove, resize, minimize, fullscreen, close windows
ocrTakes a screenshot and runs Vision OCR on it. Returns text with x/y positions. Use --store to make results clickable (O1, O2, etc.). Use when see returns no elements.
focusShow currently focused element
findSearch elements by text in latest snapshot
clipboardRead/write system clipboard
waitWait for app launch or element to appear

Always pass --json for structured output.

How to Work

You are controlling a real macOS desktop. You cannot see anything unless you explicitly look. You cannot assume anything worked unless you verify.

1. Know your environment first

Before doing anything, understand the display setup, what's running, and capture the current state of every screen:

steer screens --json       → which monitors exist, their resolution
steer apps --json          → what apps are running
steer see --screen 0 --json  → screenshot of screen 0 (primary)
steer see --screen 1 --json  → screenshot of screen 1 (if exists)

Take a screenshot of each screen returned by steer screens. Read them. This gives you a baseline of the desktop before you start making changes.

2. Focus the app, then verify

Before interacting with any app, make sure it's the active window on the right screen:

steer apps activate Safari --json
steer see --app Safari --json        → verify it's in front, read the state

3. One action, then observe

NEVER chain multiple steer commands in one bash call. The screen changes after every action. You must look after every action.

The loop is:

  1. steer see — look at the screen
  2. Read the JSON — understand what you see
  3. Do ONE action (click, type, hotkey, scroll)
  4. steer see — look again to confirm it worked
  5. Repeat

4. Clicking safely

Before clicking anything:

  • Run steer see --app <app> --json to get a fresh snapshot
  • Use element IDs from the snapshot (B1, T1, L3) — not coordinates when possible
  • If using coordinates, they are relative to the screenshot image, factor in window position
  • After clicking, run steer see again to confirm the click landed

5. Typing into fields

  • Before typing anything: ALWAYS check focus with steer focus --json to confirm the right element is targeted.
    • If you're not focused on the right element, use steer apps activate <AppName> --json to switch apps or steer click to focus the correct field.
    • Then validate your focus with steer focus --json
  • Use steer type "text" --into T1 --json to click-then-type in one step when targeting a specific field
  • If typing into the already-focused element, just steer type "text" --json
  • After typing, verify with steer see that the text appeared correctly
  • For URLs in browsers: type the URL, then steer hotkey return --json, then steer see to confirm navigation

6. Reading content from apps

Both see and ocr save a screenshot PNG. The path is in the JSON output under "screenshot". You can read this image file to visually inspect what's on screen.

Native apps (Safari, Finder, Terminal): steer see --app <name> --json gives you the accessibility tree with labeled elements. If the element list is empty, the screenshot still exists — read it or try --ocr to fall back.

Electron apps (VS Code, Slack, Notion): Accessibility trees are empty. Use OCR instead:

steer ocr --app "VS Code" --store --json

With --store, OCR results become clickable elements (O1, O2, etc.).

Web pages: Accessibility tree may be shallow. Use steer ocr --app Safari --json to read all visible text with positions. Use the positions to click precisely.

7. Waiting for things

Don't assume the UI is ready after an action. Use steer wait or run steer see in a loop:

steer wait --app Safari --json                     → wait for app to launch
steer wait --for "Submit" --app Safari --json      → wait for element to appear

8. Multi-monitor awareness

If there are multiple screens, coordinates and screenshots are per-screen. Always check steer screens --json first and use --screen <index> when needed.

Element IDs

Elements from steer see get role-based IDs: B (button), T (text field), S (static text), I (image), C (checkbox), L (link), M (menu item), O (OCR element), etc.

IDs regenerate with each snapshot. Always use IDs from the most recent steer see or steer ocr --store.

Rules

  • One command per bash call — never chain steer commands
  • Always verify — steer see after every action
  • Focus first — activate the app before interacting
  • Know your screens — check steer screens before clicking
  • Use --json always — structured output is reliable
  • Write all files to /tmp — never write output files into the project directory
  • Run steer help <cmd> if you're unsure about a command's flags

Frequently asked questions

What does the Steer AI skill do?

macOS GUI automation CLI. Use steer to see the screen, click elements, type text, send hotkeys, scroll, drag, manage windows and apps, run OCR on Electron apps, and wait for UI conditions.

Why use Steer on TypingMind?

Because you install it once and use it with any model. Steer is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Steer in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/disler/mac-mini-agent/tree/main/.claude/skills/steer. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Steer?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Steer?

As many as you like. As long as a model supports skills, you can use Steer with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Steer AI skill free?

It is published on GitHub by disler. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇