Argent Device Interaction
Drive iOS simulators, Android emulators, and Chromium apps — tap, gestures, type, screenshots
- What
- Drive iOS simulators, Android emulators, and Chromium apps — tap, gestures, type, screenshots
- Cost
- Free
- Needs
- the argent MCP server, and a target to drive: an iOS simulator, an Android emulator, or a Chromium app exposing a CDP endpoint (an Electron app or a Chromium-family browser). Physical iPhone is not supported on the interaction side, and TV targets are out of scope. If you delegate simulator tasks to sub-agents, they need MCP permissions. No account is needed to install this skill and no API key or password is involved, and nothing is sent anywhere during installation itself: installing is only copying the files below.
- Install
- Copy the installer prompt below into your Muse — your agent does the rest.
Curated by Skill Harbor — drives mobile and desktop app UIs through the argent MCP tool surface with one unified toolset: `list-devices` and `boot-device` for targets, then tap, swipe, pinch, rotate, custom gestures, scroll, type, hardware buttons, app launching, URL opening, screenshots, and wait-for-element across iOS simulators, Android emulators, and Chromium apps (Electron or a Chromium-family browser exposing a CDP endpoint — read `references/chromium.md` before driving those). Coordinates are normalized 0.0–1.0, never pixels; navigation works from the element tree returned after each action rather than screenshot pixels. Same tool names on every platform — dispatch is automatic from the target id shape. Complementary sibling: argent-metro-debugger for runtime debugging (this one is interaction, that one is inspection). Requires the argent MCP server, a simulator/emulator or a CDP-exposed Chromium app, and MCP permissions for any delegated sub-agents. Physical iPhone: not supported on the interaction side. By @software-mansion, listed here with credit to its creator. Updated October 20, 2026: upstream extended the fold tool, which folds or unfolds a foldable device by posture or angle, from foldable iOS simulators only to Android emulators as well (a Pixel Fold or generic foldable AVD), with Android-specific notes on the deviceState result, lock-screen handling after folding to closed, and the generic AVD closed-display tap limitation; the rest of the upstream change is punctuation only in SKILL.md and its three reference files, with no other change to tools or behaviour. Skill Harbor never reviews the code, review it yourself before use.
Version:
Quick answer
Argent Device Interaction is a free skill driving mobile and desktop app UIs through one unified argent MCP toolset. It covers list-devices and boot-device for targets, then tap, swipe, pinch, rotate, gestures, scroll, type, hardware buttons, app launching, URL opening, screenshots, and wait-for-element across iOS simulators, Android emulators, and Chromium apps, with coordinates normalized from zero to one, never pixels.
Install
Copy the install package below, then paste it into Muse🧪 Install ran in our test sandbox · Nov 18, 2026📌 Pinned version: commit 4304f73Step 0, before installing anything: read this install prompt, summarize in a few lines what this build does and what installing it involves, then ask me to confirm before you proceed. Prerequisites: the argent MCP server, and a target to drive: an iOS simulator, an Android emulator, or a Chromium app exposing a CDP endpoint (an Electron app or a Chromium-family browser). Physical iPhone is not supported on the interaction side, and TV targets are out of scope. If you delegate simulator tasks to sub-agents, they need MCP permissions. No account is needed to install this skill and no API key or password is involved, and nothing is sent anywhere during installation itself: installing is only copying the files below. Install "Argent Device Interaction" for me. It drives iOS simulators, Android emulators and Chromium apps through the argent MCP tool surface with one unified toolset: list-devices and boot-device for targets, then tap, swipe, pinch, rotate, custom gestures, scroll, type, hardware buttons, app launching, URL opening, screenshots, and wait-for-element, by @software-mansion, Apache-2.0 licensed. Repository: https://raw.githubusercontent.com/software-mansion/argent/4304f73208b25f1c61d476e092fdadae5692b59e/packages/skills/skills/argent-device-interact/SKILL.md (the skill is the packages/skills/skills/argent-device-interact/ folder at that commit: SKILL.md and references/chromium.md, references/gesture-examples.md and references/secrets.md) Pinned version: install the code at commit 4304f73208b25f1c61d476e092fdadae5692b59e exactly (Skill Harbor tested this version). Do not substitute a newer version. If this commit is unreachable, stop and tell me. Updated October 20, 2026: Skill Harbor regenerated this install prompt from the current upstream files. The one functional change is in the fold tool, and it is stated here plainly: fold, which folds or unfolds a foldable device by posture (closed, half-open or open) or by angle, was previously documented for a foldable iOS simulator only (the iPhone Duo); it now also covers Android emulators, a Pixel Fold or a generic foldable AVD, and list-devices marks those targets with foldable: true. The section adds two Android-specific rules: the fold result gives a deviceState (CLOSED, HALF_OPENED or OPENED), a lock screen shown by folding to closed is dismissed and a warning says when it cannot be, or when a lock screen is still up after the fold, so unlock the device before the next action; and only a Pixel Fold AVD has a cover panel, while a closed generic foldable AVD shows only the left part of its display, its describe frames do not match tap coordinates, and argent can reliably tap only an element that covers the visible part, for example a full-width row, so use a Pixel Fold AVD to test the cover panel. The two pre-existing panel rules are now labelled as iOS rules, with no change to what they say. Everything else in the diff is punctuation only: em dashes replaced by colons, semicolons or commas, and tool headings rewritten with a colon instead of a dash, in SKILL.md and in all three reference files, with no other change to tools, parameters, commands or behaviour. The interaction coverage, the coordinate system and the limits are otherwise unchanged. 1. Fetch the whole packages/skills/skills/argent-device-interact/ folder at the pinned commit, SKILL.md, references/chromium.md, references/gesture-examples.md and references/secrets.md, into a temporary folder, then compare each file with the fingerprints from Skill Harbor's review of 2026-10-20: - SKILL.md: SHA-256 0c0558b3690c65202ffd4a4599ba9b5324617dd00450e89c28eeaca6c81224e9 - references/chromium.md: SHA-256 98e3d6cb3494a34b5c1169d340c9b2bca0a376b84cadcea244d3af0c480ed204 - references/gesture-examples.md: SHA-256 ef4b6d754ee60fe6686eab841b08d59fad9b9b1d2bc56376b6b3f3c0be861fcb - references/secrets.md: SHA-256 966bcfca8b58fa036d4a32ae679d36b68953f193b665c4b721275d2eca4c81a4 If a file does not match, it changed since the review: do not install it yet, tell me what you found and let me decide. 2. Safety check: read SKILL.md and the three reference files and review them for anything suspicious (unexpected network calls, shell commands, credential harvesting); STOP on any red flag and tell me. Tell me in plain words what the skill does when used, which services it connects to, whether it reads or sends any personal information or keys, and anything that seems unusual. Confirm these points from the files: - This skill is plain-text guidance for the argent MCP tools, in four Markdown files only: SKILL.md, references/chromium.md, references/gesture-examples.md and references/secrets.md. It ships no scripts and no code of its own. Nothing is sent anywhere during installation itself; installing is only copying the folder. - In use, the agent drives a device through the argent MCP server you already run: list-devices to pick a target (results tagged ios, android or chromium, booted devices first) and boot-device when none is ready, then the interaction tools. All tools take a udid and dispatch automatically from its shape (UUID means iOS simulator, chromium-cdp-<port> means a Chromium CDP app, anything else means an Android adb serial), so the same tool names work on every platform. Gesture tool schemas may be deferred: load them with ToolSearch before the first gesture call, or parameters may be coerced to strings and fail validation. - Coordinates are normalized 0.0 to 1.0, never pixels, and navigation works from the element tree returned after each action (describe), not from screenshot pixels, except as a fallback when the tree cannot expose a reliable target. Screenshots are downscaled by default. - The fold tool now covers Android as well as iOS, as disclosed above: use it only on a foldable iOS simulator (the iPhone Duo) or Android emulator (a Pixel Fold or generic foldable AVD) marked foldable: true; other devices reject the call. Coordinates and screenshot size change with the active panel, so re-read the element tree in the fold result or run describe again before tapping, and fold between gestures, not during one. On Android, note the deviceState in the result and the lock-screen warning, and remember that only a Pixel Fold AVD has a real cover panel. - For a Chromium target, read references/chromium.md before driving it: scrolling, tabs, cookies and storage differ there. The Chromium helper tools can list, open, select and close tabs, and can read and write cookies (including HttpOnly cookies, through the CDP Network domain) and read and write local and session Web Storage for the active page and origin, which the file presents as a way to seed authentication before a flow or check app state after one. Those reads and writes happen on your own running Chromium app through its local CDP endpoint. - Typing secrets has its own rules, in references/secrets.md, to read before typing any credential: a {{secret:NAME}} placeholder in a keyboard, paste, run-sequence or flow type step is resolved on the machine running the tool-server, from an ARGENT_SECRET_<NAME> environment variable, a project .argent/secrets.env file, the ARGENT_SECRET_-prefixed keys of a project .env.local or .env, or ~/.argent/secrets.env, in that order, and nothing outside those sources is reachable. The result echoes the placeholder, never the value, and the auto-screenshot after such a call is skipped so the typed value does not re-enter the context as pixels. Never ask me to paste a secret value into the conversation; ask me to put it in a secrets file instead. - Limits and platform notes in the files: physical iPhone is not supported for interaction and TV targets are out of scope; paste works only on an iOS simulator or Android emulator; on Android, Metro needs adb reverse tcp:8081 tcp:8081 before the React Native app starts, describe fails clearly on a locked screen or secure surface, and reinstall-app takes an .apk path on Android and an .app directory on iOS. 3. When I say yes, install it as a skill: copy the whole fetched argent-device-interact folder's contents (SKILL.md and the references/ subfolder with its three files, keeping the subfolder as it is) into the agent's skills directory, in a folder named "software-mansion-argent-device-interact". 4. Verify with no network calls: frontmatter valid, files in place. 5. Report what was installed, where, and what I still need to do myself (e.g. have the argent MCP server running, and an iOS simulator, an Android emulator or a CDP-exposed Chromium app ready where the skill will run, with MCP permissions for any delegated sub-agents; no account or API key to create). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.
Saved to your recent installs. Find it anytime on /connect.
🔒 Security, provenance and what this touches
The install prompt below already includes the vetting steps: your agent follows the community checklist before installing anything with executable code. Want more?
Verification certificate
Dated facts about what was checked for this listing, and how. This is not a safety certification.
- Pinned version
- 4304f73 · 2026-10-20
- Sandbox install test
- Install ran in our test sandbox · Nov 18, 2026 · Observed during the install test: registry.npmjs.org
- Creator identity
- Not verified
- Fingerprint watch
- Last checked · Nov 19, 2026 : no change detected in the files linked by this install prompt
- Install prompt analysis
- 2026-10-13 : No commands, network calls, packages, file writes or API keys detected.
Skill Harbor never reviews the code of a listed skill. These are dated facts about the install prompt and our install test, not a safety certification. Review a skill yourself before you install it. How we verify
The Argent series
By the same creator, in the same series:
- argent-create-e2e: create, record, replay and repair reusable e2e tests
- Argent Android emulator setup: boot and connect devices for UI automation
- Argent Metro Debugger
- Argent TV — drive Apple TV, Android TV, Fire TV from an agent
- Argent iOS simulator setup
- Native mobile profiler: iOS xctrace and Android Perfetto CPU, hang and leak diagnosis
- argent-lens: propose design variants, let the human pick on device
- argent-react-native-app-workflow: start, debug and test React Native apps
- argent-react-native-optimization: measure-first RN performance pipeline
- argent-screenshot-diff: pixel-level visual regression checks
- argent-test-ui-flow: autonomous UI testing with interact-screenshot-verify loops
Questions
How do I install a skill?
Every product page includes a copy-paste install prompt. Paste it into your Muse and it sets the skill up for you — no manual configuration.
Where does my money go?
Straight to the seller. Skill Harbor never processes payments: checkout happens on the seller’s own page, usually Stripe.
What does the ✓ next to a creator’s name mean?
It means we confirmed the identity of the person behind the listing. It says nothing about the code itself — always check a skill before installing it.