software-mansion/argent
10 skills18.9k installs12.3k starsGitHub
Install
npx skills add https://github.com/software-mansion/argentSkills in this repo
1Argent Android Emulator SetupArgent Android emulator setup is an agent skill for solo and indie mobile builders who test React Native or native Android apps with Software Mansion’s Argent MCP stack. Install it when you start a fresh coding session on Android and need a reliable device serial before gestures, screenshots, or end-to-end flows. The workflow verifies SDK prerequisites, discovers ready emulators or physical devices, boots an AVD when nothing is online, and optionally reverses Metro so the emulator can reach your dev server. Hot boot targets roughly thirty seconds when snapshots work; cold boot can take several minutes, with explicit cleanup if boot stages fail. It is deliberately narrow—connection and environment—not test authoring. Pair it with argent-test-ui-flow after setup completes so your agent does not tap into a blank or offline emulator.1.9kinstalls2Argent Device InteractArgent-device-interact is a phase-specific integration skill for solo builders shipping React Native or native mobile apps who need agents to manipulate real simulators and emulators through Argent MCP. It documents a single cross-platform interaction API keyed by device id, plus the boot-and-select workflow so sub-agents with MCP permissions do not guess targets. Before gestures, you load deferred tool schemas explicitly to avoid parameter coercion bugs. Platform notes at the end cover Metro adb reverse, locked-screen describe errors, and other iOS versus Android caveats. It complements setup skills rather replacing them: list, boot, then interact. Typical use is manual-style exploratory testing, reproducing UI bugs, or capturing screenshots during build iterations—not store launch or growth analytics.1.9kinstalls3Argent Test Ui FlowArgent test UI flow teaches your coding agent to exercise real mobile UIs on iOS or Android using Software Mansion’s Argent MCP server. Solo builders install it when they need repeatable proof that login, navigation, or custom flows work on a simulator or emulator—not just unit tests on the host. The methodology is platform agnostic: the same tool names run against a UUID-style iOS udid or an adb Android serial after you resolve a device with list-devices and the appropriate setup skill. Each pass follows a disciplined loop—capture state, discover targets, interact, capture again, and verify expectations—so agents do not tap blindly. It fits Ship testing as the primary shelf but also helps Validate prototypes and Build frontend iterations before release. Always complete argent-android-emulator-setup or argent-ios-simulator-setup first so screenshots reflect a live app surface.1.9kinstalls4Argent React Native App Workflowargent-react-native-app-workflow gives solo mobile builders a disciplined way to start and debug React Native apps on iOS simulators and Android emulators when vanilla commands fail or the repo hides real entry points in custom scripts. The skill forces configuration exploration up front—typically via the argent-environment-inspector subagent—so you do not burn time on wrong Metro ports, stale Pods, or opening an .xcodeproj instead of the CocoaPods workspace. It walks through installs, pod install when needed, conflict checks, and test runs, with explicit encouragement to persist the project’s real workflow once learned. Use it whenever you are bringing up the app, fixing red screens, chasing bundle issues, or validating changes before deeper optimization or profiling work.1.9kinstalls5Argent React Native Profilerargent-react-native-profiler helps solo mobile builders quantify React Native performance on Hermes instead of guessing which screen stutters. It documents Argent profiler tools that start and stop recording sessions, capturing CPU profiles alongside React commit trees so you can see expensive renders and hotspot functions in one pass. Use it when optimizing animations or list scroll, proving a refactor helped, or producing a ranked issue report for the worst offenders. The skill explicitly pairs with argent-react-native-optimization: profiling tells you what hurts; the optimization skill carries remediation patterns. Typical flow assumes you can run the app locally—often after argent-react-native-app-workflow—then exercise the UI under load while sampling runs.1.9kinstalls6Argent Ios Simulator Setupargent-ios-simulator-setup is a short procedural agent skill for solo builders who automate iOS apps through Software Mansion’s Argent MCP stack. It walks you through finding a booted iOS simulator, calling `boot-device` when needed, and trusting that interaction tools will bring the server online. The workflow assumes you will paste or reuse UDIDs in the standard Apple format and that you may hand work to sub-agents that must already have MCP access. It does not perform gestures itself; it only establishes a reliable simulator connection at the start of a coding session or before QA-style agent runs. For indie mobile teams, this reduces failed first taps caused by cold simulators or wrong device IDs. Pair it with gesture or vision automation skills once setup completes.1.9kinstalls7Argent Metro DebuggerArgent Metro Debugger is an agent skill for solo and indie builders shipping React Native apps who need live runtime visibility instead of guess-and-redeploy loops. It walks your agent through Metro CDP prerequisites: a running Metro server, at least one connected app target, and on Android the adb reverse port forward that most connection failures trace back to. The skill catalogs how argent debugger tools accept port and logicalDeviceId so you inspect components, console logs, and evaluated JavaScript on the intended device. That matters when you run two simulators on the same port or debug physical Android hardware. Use it during build iterations and again in ship when you chase flaky UI or network behavior in a dev build. It complements argent-create-flow for scripted taps but stays focused on interactive CDP debugging rather than recorded replay.1.9kinstalls8Argent React Native Optimizationargent-react-native-optimization is an agent skill that turns React Native performance investigation into actionable code changes. Solo mobile builders use it after profiling shows repeated renders, unstable props, inline style objects, or list scroll jank. The skill is structured as a finding-to-fix table: each row names a symptom, the React API to apply, and guardrails such as deferring React.memo when the React Compiler already optimizes a component. It complements mansion-style profiler workflows by keeping decisions consistent—when to memoize children, when module-level constants beat inline arrays, and when FlashList replaces tuned FlatList props. It does not replace measuring on device; it accelerates the edit pass once you know what the trace implicates. Intermediate to advanced RN experience helps because wrong memo boundaries can hide bugs.1.9kinstalls9Argent Create FlowArgent Create Flow is an agent skill for solo mobile builders using Software Mansion's argent stack who tire of re-typing the same tap, scroll, and debugger sequences. It documents a record-verify-save loop: start recording with a name and executionPrerequisite, add steps that execute MCP tools immediately and append to YAML only on success, sprinkle echo labels for readable replays, then finish and execute by flow name. Flows live under .argent/flows/, which keeps automation versionable next to the repo. This is distinct from argent-metro-debugger's interactive CDP work—create-flow optimizes repeatable device choreography for regression smoke paths or demo scripts. Intermediate complexity assumes you already invoke argent MCP tools correctly; the skill enforces learning-by-doing because broken steps never enter the file. After a flow exists, hand off to CI or manual flow-execute before ship checkpoints.1.9kinstalls10Argent Native ProfilerArgent Native Profiler is an agent skill for solo mobile builders diagnosing performance below the JavaScript layer. It documents a small toolkit to start and stop Xcode Instruments-style recordings, export timestamped trace XML, parse bottlenecks, and query stacks for hangs, CPU hotspots, and leaks. Today the backend is iOS-only through xctrace with Xcode CLI tools on PATH; Android serials fail fast with a clear roadmap message. The skill emphasizes repeatable investigation patterns after analyze surfaces issues, so you move from symptom to native call chains instead of guessing in Metro logs. It suits indie React Native teams shipping performance-sensitive apps who need native evidence before rewriting UI or bridge code.1.9kinstalls