agentsclimarketplace

Marionette

Skill Poorgramer-Zack/dart-expert-skills/skills/marionette

A comprehensive library of modular Agent Skills for Flutter & Dart development

Install
npx -y skills add Poorgramer-Zack/dart-expert-skills --skill marionette

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

AI-driven Flutter E2E testing via VM Service CLI. Control running Flutter apps in debug mode for smoke tests, regression checks, exploratory testing, and UI automation. Trigger when user mentions E2E testing, UI automation, smoke tests, integration testing, test automation, controlling Flutter apps, VM Service interaction, debugging UI flows, automated regression testing, CI sanity checks, or testing native dialogs/permissions/WebViews. Supports stateful named instances or stateless direct URI connections. Ideal for rapid test feedback, AI agent workflows, element discovery, screenshot capture, log collection, and scriptable UI interactions (tap, scroll, enterText). Use for troubleshooting flaky tests, automating repetitive manual tests, verifying app behavior after merges, or setting up automated QA pipelines.

SKILL.md

8.1 KB, ~1.9k tokens by cl100k_base, as published. Nobody here has run it

Marionette CLI — AI Agent Reference

Marionette CLI controls Flutter apps running in debug mode. It supports multiple simultaneous app instances via a named instance registry, or direct URI connections for fully stateless operation.

Workflow

Option A: Named Instances (stateful)

  1. Start your Flutter app(s) in debug mode and note VM service URI(s) (printed in console, e.g., ws://127.0.0.1:XXXXX/ws).
  2. Register each app: marionette register <name> <uri>
  3. Interact using: marionette -i <name> <command> [args]
  4. Clean up when done: marionette unregister <name>

Option B: Direct URI (stateless)

  1. Start your Flutter app in debug mode and note the VM service URI.
  2. Interact directly: marionette --uri <ws-uri> <command> [args]

No registration, no cleanup, no files on disk. Each command opens a fresh WebSocket connection, executes, and disconnects.

Global Options

-i, --instance <name> Target instance (required unless --uri is used) --uri <ws-uri> VM service WebSocket URI — bypasses registry, mutually exclusive with --instance --timeout <seconds> Connection timeout (default: 5)

Commands

register <name> <uri>

Register a Flutter app instance.

Arguments: name Alphanumeric identifier [a-zA-Z0-9_-]+ uri VM service WebSocket URI (e.g., ws://127.0.0.1:8181/ws)

Example: marionette register my-app ws://127.0.0.1:8181/ws

Output (stdout): Registered instance "my-app" → ws://127.0.0.1:8181/ws

Output if overwriting (stderr): Updated existing instance "my-app" → ws://127.0.0.1:8181/ws

Exit codes: 0 success, 64 invalid name/usage


unregister <name>

Remove a registered instance.

Arguments: name Instance name to remove

Example: marionette unregister my-app

Output (stdout): Unregistered instance "my-app".

Output if not found (stderr, exit 1): Instance "my-app" not found.


list

List all registered instances.

Example: marionette list

Output (stdout): Registered instances:

  my-app
    URI: ws://127.0.0.1:8181/ws
    Registered: 2026-02-12 15:30:00.000

Output if empty (stdout): No instances registered.


get-interactive-elements

List interactive UI elements in the app's widget tree.

Requires: -i <instance> or --uri <ws-uri>

Examples: marionette -i my-app get-interactive-elements marionette --uri ws://127.0.0.1:8181/ws get-interactive-elements

Output (stdout), one line per element: Found 3 interactive element(s):

Type: ElevatedButton, Key: "submit_button", Text: "Submit"
Type: TextField, Key: "email_field"
Type: IconButton, Text: ""

Each element may have: type, key, text, and additional properties. Use the key or text values as matchers for tap, enter-text, scroll-to.


tap

Tap an element. Provide exactly one matching strategy.

Requires: -i <instance> or --uri <ws-uri>

Options: --key <string> Match by ValueKey<String> (most reliable) --text <string> Match by visible text content --type <string> Match by widget type name (e.g., ElevatedButton) --x <number> X screen coordinate (use with --y) --y <number> Y screen coordinate (use with --x)

Examples: marionette -i my-app tap --key submit_button marionette -i my-app tap --text "Submit" marionette --uri ws://127.0.0.1:8181/ws tap --key submit_button marionette -i my-app tap --x 100 --y 200

Output (stdout): Tapped element matching {key: submit_button}


enter-text

Enter text into a text field.

Requires: -i <instance> or --uri <ws-uri>

Options (all required): --key <string> Match text field by key (or use --text) --text <string> Match text field by visible text --input <string> Text to enter (mandatory)

Example: marionette -i my-app enter-text --key email_field --input "[email protected]"

Output (stdout): Entered text into element matching {key: email_field}


scroll-to

Scroll until an element becomes visible.

Requires: -i <instance> or --uri <ws-uri>

Options: --key <string> Match by ValueKey<String> --text <string> Match by visible text content

Example: marionette -i my-app scroll-to --text "Bottom Item"

Output (stdout): Scrolled to element matching {text: Bottom Item}


take-screenshots

Capture screenshots and save to PNG files.

Requires: -i <instance> or --uri <ws-uri>

Options: -o, --output <path> Output file path (mandatory) --open Open the file after saving

Example: marionette -i my-app take-screenshots --output ./screenshot.png

Output (stdout): Saved screenshot: ./screenshot.png

Multi-view apps produce numbered files: Saved screenshot: ./screenshot.png Saved screenshot: ./screenshot_1.png


get-logs

Retrieve collected application logs.

Requires: -i <instance> or --uri <ws-uri>

Example: marionette -i my-app get-logs

Output (stdout): Collected 5 log entries:

[INFO] App started
[DEBUG] Loading data...
...

Output if empty (stdout): No logs collected.


hot-reload

Perform a hot reload of the Flutter app.

Requires: -i <instance> or --uri <ws-uri>

Example: marionette -i my-app hot-reload

Output (stdout, exit 0): Hot reload completed successfully.

Output on failure (stderr, exit 1): Hot reload failed. The app may need a full restart.


doctor

Check connectivity of all registered instances.

Example: marionette doctor

Output (stdout): Checking 2 instance(s)...

  my-app (ws://127.0.0.1:8181/ws) ... OK
  other-app (ws://127.0.0.1:9090/ws) ... FAILED

Some instances are unreachable. Use "marionette unregister <name>" to remove stale entries.

Exit codes: 0 all reachable, 1 any unreachable


mcp

Run the Marionette MCP server (preserves original marionette_mcp behavior).

Options: -l, --log-level <level> FINEST|FINER|FINE|CONFIG|INFO|WARNING|SEVERE (default: INFO) --log-file <path> Log file path (default: stderr) --sse-port <port> Use SSE transport on this port (default: stdio)

Example: marionette mcp marionette mcp --sse-port 3000


Exit Codes

0 Success 1 Runtime error (connection failed, command failed, app unreachable) 64 Usage error (missing arguments, invalid options)

Error Recovery

If a command fails with a connection error, the app may have stopped.

  • --instance mode: Run marionette doctor to check all instances, then marionette unregister <name> to clean up stale entries.
  • --uri mode: Verify the URI is correct and the app is still running. Re-run flutter run if needed and use the new URI.

Tips

  • Prefer --uri for one-off interactions (no setup/cleanup overhead)
  • Prefer --instance for repeated interactions with the same app (shorter commands)
  • Prefer --key over --text for matching elements (keys are stable, text may change)
  • Run get-interactive-elements first to discover what's on screen before interacting
  • Instance names are alphanumeric with hyphens/underscores: [a-zA-Z0-9_-]+
  • Commands are stateless — each opens a fresh connection, so no session management needed

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.