Phone Use logo

Phone Use

Community
niki914
Phone Use

MUST Load this skill for ANY task that involves GUI interaction or device control — opening apps, tapping, typing, scrolling, swiping, Back/Home/Recents, reading the screen, or any workflow that touches on-screen content. Load this even when the user does not explicitly mention the device — if the task implies a GUI operation, load this skill first.

Overview

Publisherniki914
Repositoryzafiro
Skill namePhone Use
Stars
305
Forks
50
Bundled files
Instructions only
LicenseMIT
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 niki914 on GitHub. Read the source before you install it.

Installation

Install the Phone Use 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/niki914/zafiro/tree/main/app/src/main/assets/skills/phone-use
  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/niki914/zafiro.git /tmp/zafiro
mkdir -p .claude/skills
cp -r /tmp/zafiro/app/src/main/assets/skills/phone-use .claude/skills/niki914-phone-use
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Phone Use 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 Phone Use 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 Phone Use 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.

Overview

You can see and control the Android screen through three channels:

  • The UI tree — screen_operation_accessibility reads the semantic tree (text, labels, bounds, clickability) and acts on nodes by token.
  • Pixels — screenshot captures the screen as an image and view_image reads image files, so you can visually inspect any app.
  • Coordinates — screen_operation_shell injects tap/swipe/key events at screen-pixel coordinates.

Use the tree first; drop to pixels and coordinates when the tree cannot see or reach the target. Coordinates may come from node bounds or from a screenshot — compute them, never guess.

First-use consent: when the accessibility or overlay permission is missing, the first screen tool call asks the user to grant both. If the user declines, the tool fails with a consent message — tell the user what to enable and stop; do not retry in a loop. Root is not required; it only shortens the permission chain.

Workflow

1. Open the target

Prefer open_uri when a deep link lands closer to the goal than the launcher activity (e.g. a map at coordinates, a settings page). Otherwise resolve with find_installed_apps if the name is ambiguous, then launch_app.

2. Read the screen

Start with screen_operation_accessibility(operation: "read"). If the tree is empty or root-only, the app uses a non-native UI (Flutter, Unity, WebView, self-drawn canvas) — the tree will not work for this app. Do not retry the read. Instead switch to screenshot + coordinate operations via screen_operation_shell.

3. Act

  • Tree shows a labeled, clickable node → tap it by token.
  • Target is unlabeled, a bare container, or the node's bounds look misaligned with what you see → take a screenshot, locate the real target visually, convert image coordinates to screen pixels (see below), and tap with screen_operation_shell.
  • set_text needs the accessibility method; there is no shell fallback for typing.

Image → screen coordinates. The screenshot is downscaled; its width/height in the result are the image's, not the screen's. Screen dimensions are in the tree header. Scale per axis: screen_x = image_x * screen_w / image_w (same for y).

4. Verify

Write operations return the updated tree automatically. Check that the UI actually reached the expected state — a success status means the event was injected, not that the app reacted. If the screen did not change as expected, re-examine: the target may be self-drawn (bounds unreliable — see §3), the tap may have landed on an overlay, or the app may need more time (wait_ms). On SHELL_TIMEOUT or SHELL_SESSION_LOST, verify the screen before retrying as the action may have partially executed.

5. Scroll

Scroll lists with screen_operation_shell(operation: "swipe", ...) — accessibility scroll steps are app-defined and unreliable. Derive coordinates from the scrollable container's bounds, swiping ~70% of the container. Start the gesture mid-container, not at the bottom edge: bottom-docked bars (tab bars, mini-players, toolbars) steal the touch-down. If a dock's bounds contain the start point, move the start above its top edge.

6. Navigate

Use key events via screen_operation_shell: Back (code: 4), Home (code: 3), Recents (code: 187).

Scrolling heuristics

  • Before scrolling a list, define the exit condition in observable UI: a result count, an empty state, a section boundary.
  • After two separate swipes the visible items are unchanged and nothing is loading → the list boundary is reached. Retarget once in case the gesture missed the container before concluding.
  • One swipe per decision when hunting a specific item; batching 2–3 swipes is acceptable only when skipping intermediate states cannot miss the target.

App traps

When an unexpected result reveals a stable, verified rule about an app's UI, save it with memory (one natural-language sentence naming the app, the misleading element, and the correct action). Do not memorize temporary state, the current tab, or unverified guesses.

Frequently asked questions

What does the Phone Use AI skill do?

MUST Load this skill for ANY task that involves GUI interaction or device control — opening apps, tapping, typing, scrolling, swiping, Back/Home/Recents, reading the screen, or any workflow that touches on-screen content. Load this even when the user does not explicitly mention the device — if the task implies a GUI operation, load this skill first.

Why use Phone Use on TypingMind?

Because you install it once and use it with any model. Phone Use 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 Phone Use in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/niki914/zafiro/tree/main/app/src/main/assets/skills/phone-use. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Phone Use?

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 Phone Use?

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

Is the Phone Use AI skill free?

Yes. It is published on GitHub by niki914 under the MIT license. 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 👇