Flutter Map logo

Flutter Map

Community
tayormi
flutter-map

Generate a visual navigation map of a Flutter app. Statically parses GoRouter / go_router_builder / AutoRoute / Navigator 1.0 / Navigator 2.0 into expo-map-compatible graph.json, prepares deep-link explore flows, then deep-links through every screen on iOS Simulator or Android emulator capturing screenshots and runtime states, and packs a portable .appmap bundle. Use when the user asks to map a Flutter app's navigation, screens, or routes, or wants a visual sitemap of their Flutter app.

Overview

Publishertayormi
Repositoryflutter-map
Skill nameflutter-map
Stars
90
Forks
7
Bundled files
Instructions only
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 tayormi on GitHub. Read the source before you install it.

Installation

Install the Flutter Map 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/tayormi/flutter-map/tree/main/skills/flutter-map
  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/tayormi/flutter-map.git /tmp/flutter-map
mkdir -p .claude/skills
cp -r /tmp/flutter-map/skills/flutter-map .claude/skills/flutter-map
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Flutter Map 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 Flutter Map 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 Flutter Map 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.

flutter-map

Produce a visual map of a Flutter app's navigation: every route as a card with a screenshot, runtime state variants (bottom sheets, dialogs, modals), and navigation edges — packed into an expo-map-compatible .appmap bundle.

Arguments: optional path to the Flutter project (default: cwd). --static = parse + pack only (no simulator). --prepare = parse + write deep-link flow stubs / explore plan, stop before live capture.

Working directory: all outputs go to <project>/.flutter-map/ — graph.json, explore-plan.json, capture-status.json, screens/*.png, flows/*.yaml + flows/*.meta.json, *.appmap. Suggest adding .flutter-map/ to .gitignore.

Parser / pack CLIs live in this repo at packages/flutter_map_parser:

bash
PARSER=<flutter-map-repo>/packages/flutter_map_parser
dart run $PARSER/bin/parse_routes.dart <project>
dart run $PARSER/bin/prepare_explore.dart <project>
dart run $PARSER/bin/pack_map.dart <project>
# or: dart run $PARSER/bin/flutter_map.dart <project>

Flow recording (Phases 4–5)

Record interactions as argent flows: <name>.yaml + <name>.meta.json under .flutter-map/flows/. Replay: npx @swmansion/argent flow run .flutter-map/flows/<name>.yaml. Schema: docs/appmap-format.md.

Rules:

  • Prefer argent MCP recording when connected; otherwise write YAML with normalized 0–1 coordinates.
  • Sidecar steps keyed by 0-based YAML index. Navigating taps need "screen": "<route id>". Screenshots are sidecar capture fields — never YAML steps.
  • Record each successful action when it happens; do not reconstruct flows afterward. Prefer durable accessibility/text selectors over coordinates whenever Argent exposes them.
  • Attach capture to the YAML step after which the screenshot was taken. Every coordinate tap/swipe also needs a human-readable target or note.
  • Mechanical deep-link flows are generated by prepare_explore — keep and update them after live captures.
  • Never record credentials; use {{secret:NAME}} if input is unavoidable.
  • Never tap destructive controls (delete, purchase, sign out, submit).

Phase 1 — static parse

bash
dart run $PARSER/bin/parse_routes.dart <project>

Report graph.json summary: mode (go_router / go_router_builder / auto_route / navigator / navigator_2), route count, edges (flag unresolved), state hints, params, scheme. If mode is unknown, stop.

Edges resolve literal paths, path helpers/consts (AppPaths.foo, Routes.settings.root, getters like Routes.newPersonPath), and interpolated literals ('${Routes.tutorials}/$id'). AutoRoute also covers context.router.push, context.pushRoute, AutoRouter.of(context).push/replaceAll, router fields like _appRouter?.push, and Navigator.push of known *Screen/*Page widgets. Navigator mode also covers pushNamed / Navigator.pushNamed. Navigator 2 mode derives routes from RouterDelegate pages: stacks (plus ValueKey / RouteInformationParser path hints) and emits root→page stack edges. Shared widgets may still attribute from imperfectly.

If --static, jump to Phase 6 (pack with missing captures).

Phase 2 — prepare explore plan

bash
dart run $PARSER/bin/prepare_explore.dart <project>

This writes:

  • .flutter-map/explore-plan.json — per-route deep links, sample params, simctl/adb commands, screenshot paths
  • .flutter-map/flows/deeplink-<slug>.yaml + .meta.json — deterministic open-url flows
  • .flutter-map/capture-status.json — all routes start as missing
  • .flutter-map/screens/ directory

Review sample params: prefer values from resolved edge targets (e.g. /details/42), then fixture/seed data, else 1. Edit explore-plan.json if needed before sweeping. Auth-gated routes and redirects are expected findings: plan how to distinguish a real auth wall from a broken deep link before capture begins.

If --prepare, stop here and report the plan.

Phase 3 — boot the Flutter app

  1. Prefer an already-running app on a booted simulator/emulator.
  2. iOS: xcrun simctl list devices booted. Android: adb devices.
  3. Run the app (flutter run -d <device> or attach to an existing debug session). Note the bundle id / application id.
  4. Verify deep linking using the scheme from graph.json / explore-plan.json:
    • iOS: xcrun simctl openurl booted "<scheme>://"
    • Android: adb shell am start -a android.intent.action.VIEW -d "<scheme>://"
  5. Confirm the app rendered (screenshot). If deep links fail, fix Android intent-filters / iOS URL types, or continue with in-app navigation only and mark needsNavigation later.

Phase 4 — route sweep

For each route in explore-plan.json:

  1. Open deepLink (or the platform command in the plan).
  2. Wait ~1.5s (3–4s for network-heavy screens).
  3. Capture using the exact slug:
    • iOS: xcrun simctl io booted screenshot <project>/.flutter-map/screens/<slug>.png
    • Android: adb exec-out screencap -p > <project>/.flutter-map/screens/<slug>.png
  4. Update the matching flows/deeplink-<slug>.meta.json device field once known.
  5. Sanity-check periodically. Keep bad captures but note them for Phase 4b.

Phase 4b — review and recover

Render and open a draft contact map before classifying captures:

bash
dart run $PARSER/bin/render_map.dart <project>

Inspect every card. Loading skeletons get a longer retry; fake IDs should be replaced with real fixture/API values; redirects are classified by what actually rendered. After any red error screen or sticky error boundary, restart the app and re-capture the remaining tail so one failure does not silently poison later screenshots.

Write verdicts to .flutter-map/capture-status.json:

json
{ "<routeId>": { "status": "ok", "needsNavigation": false, "note": null } }

Statuses: ok · empty-state · not-found · error-boundary · loading · auth-wall · missing.

Flutter-specific notes:

  • Red error screens / yellow-black stripes → error-boundary; relaunch the app before continuing (flutter hot restart, or terminate + relaunch) — error UI can poison later captures.
  • GoRouter redirects (auth) → capture what rendered; set auth-wall or note the redirect target.
  • Routes that only work via typed context.router.push / shell tabs may need needsNavigation: true if bare deep links land on the wrong stack.

Phase 5 — runtime states

For each route with non-empty stateHints:

  • bottom-sheet — find the trigger in file, tap it, capture screens/<slug>--sheet.png (or --sheet-<snap> if snap points exist).
  • dialog — trigger, capture screens/<slug>--dialog.png, dismiss.
  • modal / router-modal — capture variant or treat as own route if already in the graph.

Coverage line required: state hints: N · captured M · skipped K (why). Treat the hint list as a required work queue. Re-open the base route between states, capture every parsed sheet snap point, and explicitly record why any hint could not be triggered safely.

Phase 5b — navigation flows

For every reachable route, record flows/nav-<slug>.yaml + sidecar from app root. Walk the app depth-first and reuse verified prefixes, while keeping each flow self-contained. Every navigating tap/swipe must include the destination screen after visually confirming the landing.

When an action opens a drawer, menu, sheet, or dialog without changing routes, capture it once as <source-slug>--<state>.png and attach it to that action in every flow using it. Screens with no safe discoverable path should be marked needsNavigation with a concrete reason instead of inventing a flow.

Phase 6 — pack and deliver

bash
# optional downscale on macOS
for f in <project>/.flutter-map/screens/*.png; do sips -Z 800 "$f" >/dev/null; done

dart run $PARSER/bin/pack_map.dart <project>
dart run $PARSER/bin/render_map.dart <project> # self-contained .flutter-map/map.html

Deliver the .appmap zip. Open it in apps/flutter_map_visualiser (flutter run -d chrome) or the expo-map visualiser (drag-drop). Report: routes captured / total, state variants, skips, unresolved edges. Suggest gitignoring .flutter-map/.

Replay

bash
npx @swmansion/argent flow run <project>/.flutter-map/flows/<flow-name>.yaml

Fallback: execute YAML steps via simctl/adb + simulator MCP using sidecar target labels.

Legacy JSON flows can be migrated once with:

bash
dart run $PARSER/bin/convert_flows.dart <project> --delete-v1

Web fallback

When no simulator/emulator is available, run the app in Chrome, enumerate the same parsed paths, and capture each route at a fixed phone-sized viewport. Browser capture cannot validate native navigation stacks or system gestures, so report those limitations and do not claim navigation-flow coverage unless the interactions were actually replayed and verified.

Frequently asked questions

What does the Flutter Map AI skill do?

Generate a visual navigation map of a Flutter app. Statically parses GoRouter / go_router_builder / AutoRoute / Navigator 1.0 / Navigator 2.0 into expo-map-compatible graph.json, prepares deep-link explore flows, then deep-links through every screen on iOS Simulator or Android emulator capturing screenshots and runtime states, and packs a portable .appmap bundle. Use when the user asks to map a Flutter app's navigation, screens, or routes, or wants a visual sitemap of their Flutter app.

Why use Flutter Map on TypingMind?

Because you install it once and use it with any model. Flutter Map 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 Flutter Map in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/tayormi/flutter-map/tree/main/skills/flutter-map. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Flutter Map?

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 Flutter Map?

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

Is the Flutter Map AI skill free?

It is published on GitHub by tayormi. Check the repository for licensing terms. 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 👇