1. Habit
  2. Skills
  3. agent-device skill
callstack · 18K installs

agent-device skill
Runs by itself, every time a new app build is ready.

Every new build gets a real test.

Install the agent-device skill and make it a habit: follow https://gethabit.dev/skills/callstack/agent-device.md

  1. 1Copy
  2. 2Paste into Claude Code, Codex or Cursor
  3. 3It runs by itself

18K installs5K starsMIT licenceSource on GitHub ↗

agent-device

by callstack · 18K installs

The skill

Taps through your app on a phone or TV like a real user.

Only when you ask.

Fewer bugs

Every new build gets a real test. Broken screens are found the day they appear.

↻ Runs by itselfevery time a new app build is ready

How the habit works.
You don't have to.

Three questionsWhen does it start? How does it check its work? What does it remember?

01

It starts by itself

new app build ready. Runs in GitHub Actions or Claude Code routine.

  • ↻ new app build ready
  • build finished on main

02

It checks its work

expected text on screen. If a check fails, it tries again.

  • ✓agent-device open returns a snapshot with refs
  • ✓wait text finds the expected text
  • ✓is or get confirms the named expectation
  • ✓agent-device close ends the session

03

It remembers

flow results per build. The next run reads it first: flows that failed on the last build.

loops/app-smoke-flow-loop/run-log.md

One run, step by step.
Then it stops.

Stop ruleStops after 3 flows per build or 1 retry per failed flow.

  1. 01

    Open the app

    Your agent runs agent-device open with --foreground and gets the first snapshot with refs.

  2. 02

    Drive the flow

    It uses press, fill or scroll with --settle and follows the printed diff.

  3. 03

    Reach far targets

    For items off screen it uses scroll down --until with a selector.

  4. 04

    Check the result

    It confirms the expected screen with wait text, is, get or find.

  5. 05

    Close and log

    It runs agent-device close and writes pass or fail for the flow.

What it may do.
Nothing more.

Safe by defaultIt writes only in its own folder and asks you before anything goes out.

Is it for you?
Use it when it fits.

Quick checkRead both lists. If the left one sounds like you, the agent-device skill helps.

Use it when

  • ✓ You test an app flow on a device
  • ✓ You need screenshots or UI info
  • ✓ You drive a TV app remote

Skip it when

  • – Web pages in a browser
  • – Pure unit tests without a device

Questions

agent-device skill, answered.

Which devices does it work on?

The skill covers iOS, tvOS, macOS, Android and Amazon Vega OS TV apps in Vega Virtual Devices.

What if the snapshot has no usable refs?

The skill falls back to a screenshot and coordinates, then tries snapshot -i again after it navigates.

How do I install the agent-device skill?

Copy the prompt at the top into your agent. It installs the skill from callstack/agent-device and sets up the habit. Or run npx skills add callstack/agent-device --skill agent-device.

Is the agent-device skill free?

Yes. It is published under MIT. Habit is free too.

agent-device. Now as a habit.

Install the agent-device skill and make it a habit: follow https://gethabit.dev/skills/callstack/agent-device.md

  • Free
  • No account
  • Undo in one line

All callstack skills →More code skills →

Works in

  • Claude Code
  • Codex
  • Cursor
  • ChatGPT
  • Claude app

Made by callstack, published under MIT. Habit links to the original and never changes it. Is this your skill?

Copying didn't work here. Select the line and copy it by hand.