mobile-ai-agents 1.0.23

package/docs/roadmap/github-issues.md

@@ -0,0 +1,239 @@
+# GitHub Roadmap Issues
+
+These are the 10 planned issues for the v0.2.0 milestone. Copy each into a GitHub Issue directly.
+
+---
+
+## Issue 1
+
+**Title:** `[AGENT] Android APK/AAB Size Analyzer`  
+**Labels:** `new-agent`, `android`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add a new agent that reviews an Android project's build output and Gradle configuration to identify APK/AAB size reduction opportunities.
+
+**What it should analyze:**
+- `build/outputs/mapping/` — detect unused code that R8 is not shrinking
+- `build/outputs/apk/` — identify large resource files (images over 100KB, raw assets)
+- `build.gradle.kts` — detect missing `isShrinkResources`, disabled minification, missing `splits` config
+- `libs.versions.toml` — flag dependencies known for large transitive size (e.g., full Guava instead of the Android-safe subset)
+- ProGuard/R8 keep rules — over-broad `-keep` rules that prevent code shrinking
+
+**Input format:**
+- APK analysis output (from `bundletool` or Android Studio's APK Analyzer)
+- `build.gradle.kts` content
+- Optional: `merged-manifest.xml`
+
+**Expected output:**
+- Ranked list of size reduction opportunities with estimated savings per item
+- Corrected Gradle config for each optimization
+- ProGuard rule improvements
+
+**Reference:** Android docs on [Reduce your app size](https://developer.android.com/topic/performance/reduce-apk-size)
+
+---
+
+## Issue 2
+
+**Title:** `[AGENT] iOS App Store Review Guidelines Compliance Checker`  
+**Labels:** `new-agent`, `ios`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add an agent that reviews an iOS app's code and `Info.plist` against Apple's App Store Review Guidelines to flag likely rejection reasons before submission.
+
+**What it should check:**
+- Privacy: `NSLocationWhenInUseUsageDescription` present when `CLLocationManager` is used; no undeclared data collection
+- Permissions: No permissions requested without clear user-facing justification in usage strings
+- In-app purchases: No external payment links (violates guideline 3.1.1)
+- Age rating: Flag adult content, violence, gambling patterns without declared rating
+- Deep links: No URL schemes that conflict with system apps
+- Hidden features: No disabled-by-flag code that activates remotely (guideline 2.5.2)
+- Login requirements: Non-account features gated behind login (guideline 5.1.1)
+
+**Input format:** `Info.plist` content + Swift/ObjC source files
+
+**Expected output:** Numbered list of potential guideline violations with the specific guideline reference (e.g., 2.1 App Completeness), severity, and corrective action.
+
+---
+
+## Issue 3
+
+**Title:** `[AGENT] Flutter State Architecture Advisor`  
+**Labels:** `new-agent`, `flutter`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add an agent that analyzes a Flutter project description and recommends the right state management approach, with a migration path if they're currently using the wrong one.
+
+**Approach matrix to cover:**
+
+| Scenario | Recommendation |
+|---|---|
+| Solo dev, small app, no team | `setState` + `ValueNotifier` |
+| Team, complex features, testability required | Riverpod (code gen) or BLoC |
+| Real-time data (WebSocket, Firebase) | BLoC with `StreamSubscription` |
+| Shared state across many unrelated widgets | Riverpod |
+| Feature-scoped state, no global sharing | Cubit |
+| Large team, need strict conventions | BLoC (events enforced) |
+
+**Input:** Project description, team size, feature complexity, current state management  
+**Output:** Recommendation with rationale, migration plan if switching, and 3 code examples
+
+---
+
+## Issue 4
+
+**Title:** `[INTEGRATION] Windsurf IDE Integration Guide`  
+**Labels:** `documentation`, `integration`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add `docs/windsurf-integration.md` explaining how to use agents in Windsurf (Codeium's AI IDE).
+
+**Content to cover:**
+- How Windsurf reads project-level AI instructions (rules files)
+- How to reference agent files in Cascade (Windsurf's AI chat)
+- How to use Windsurf's multi-file context with agents that review multiple files
+- Comparison with Cursor workflow
+- Example session: open a Flutter BLoC feature, run the BLoC Feature Builder agent, apply the output
+
+---
+
+## Issue 5
+
+**Title:** `[AGENT] React Native New Architecture Compatibility Auditor`  
+**Labels:** `new-agent`, `react-native`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add an agent that audits a React Native project for New Architecture (JSI/Fabric/TurboModules) compatibility — beyond the existing migration guide prompt.
+
+**What it should audit:**
+- `package.json` — flag community libraries with no New Arch support (check community directory)
+- Native modules — identify which use legacy `ReactContextBaseJavaModule` (Android) or `RCTBridgeModule` (iOS) and need TurboModule migration
+- Native UI components — identify `requireNativeComponent` calls that need Fabric migration
+- `metro.config.js` — check for correct New Arch configuration
+- `gradle.properties` / `Podfile` — verify New Arch enablement flags
+
+**Input:** `package.json`, native module list, and RN version  
+**Output:** Compatibility report sorted by blocking vs. non-blocking issues, with effort estimates per migration item
+
+---
+
+## Issue 6
+
+**Title:** `[AGENT] Unity Mobile Performance Profiler Advisor`  
+**Labels:** `new-agent`, `unity`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add an agent that analyzes Unity Profiler output (frame data, memory snapshots) for mobile-specific performance issues.
+
+**What it should identify:**
+- Draw calls above the mobile budget (>100 on mid-range devices)
+- GC.Alloc spikes in hot paths (Update, coroutines, LINQ in hot code)
+- Texture memory overuse (>150MB VRAM on mobile)
+- Physics.FixedUpdate taking >2ms per frame
+- Shader compilation hitches on first render
+- Audio mixer complexity causing CPU overhead
+
+**Input:** Unity Profiler JSON export or frame time summary paste  
+**Output:** Ranked findings with specific Unity API or settings change for each
+
+---
+
+## Issue 7
+
+**Title:** `[PROMPT] Android Jetpack Compose Animation Prompt Pack`  
+**Labels:** `new-prompt`, `android`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add a prompt file at `prompts/android/compose-animations.md` covering the most common Compose animation needs:
+
+- `AnimatedVisibility` with custom enter/exit transitions
+- `animateContentSize` for expandable cards
+- Shared element transitions (Compose 1.7+)
+- `Crossfade` between loading/content states
+- `AnimatedCounter` (number counting up animation)
+- `ShimmerEffect` loading placeholder
+
+Each prompt should produce a complete, self-contained Composable with a `@Preview`.
+
+---
+
+## Issue 8
+
+**Title:** `[SKILL] Android Performance Skill Module`  
+**Labels:** `new-skill`, `android`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add `skills/android/performance.md` covering Android-specific performance patterns beyond what the existing Compose reviewer covers.
+
+**Content to include:**
+- Main thread rules: no disk I/O, no network, no heavy computation
+- RecyclerView optimization: `DiffUtil`, `setHasFixedSize`, `setItemViewCacheSize`
+- Bitmap handling: `BitmapFactory.Options.inSampleSize`, `inBitmap` reuse
+- Startup performance: `App Startup library`, lazy initialization, splash screen best practices
+- Tracing: `Trace.beginSection()` for custom Perfetto traces
+- Baseline Profiles: when and how to generate them
+
+---
+
+## Issue 9
+
+**Title:** `[WORKFLOW] Multi-File Feature Review Workflow`  
+**Labels:** `new-agent`, `cross-platform`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add a workflow agent at `agents/cross-platform/feature-reviewer/` that reviews an entire feature folder (multiple files) as a coherent unit, rather than one file at a time.
+
+**Why this matters:** Individual file reviewers miss cross-file issues:
+- ViewModel calling Repository directly (should go through UseCase)
+- Fragment receiving a Domain entity (should receive a UI model)
+- Use case importing Android framework classes (Clean Architecture violation)
+- Repository method naming inconsistency across files
+
+**Input:** Paste multiple files with `--- FILE: path ---` separators  
+**Output:** Cross-file findings that cannot be caught in per-file review, plus a per-file summary
+
+---
+
+## Issue 10
+
+**Title:** `[DOCS] Video Walkthrough — Using Agents in a Real Android Project`  
+**Labels:** `documentation`, `community`, `enhancement`  
+**Milestone:** v0.2.0
+
+### Description
+
+Add `docs/video-walkthroughs.md` linking to community-contributed screencasts showing agents used in real projects.
+
+**Format for each entry:**
+```
+### [Agent Name] in a [Project Type]
+- **Contributor:** @username
+- **Duration:** X minutes
+- **What it covers:** brief description
+- **Link:** [YouTube / Loom]
+```
+
+**Seed content (to be added):**
+- Crash analyzing a production ANR in a ride-sharing app
+- Generating a complete BLoC feature in a Flutter e-commerce app
+- Using the Compose Screen Builder to scaffold an onboarding flow
+
+This issue is a call for community contributions — add your own walkthrough by opening a PR to `docs/video-walkthroughs.md`.

package/docs/skill-categories.md

@@ -0,0 +1,125 @@
+# Skill Categories
+
+This document describes the taxonomy used across agents, skills, and prompts in this repository.
+
+---
+
+## Core Loop Taxonomy
+
+Mobile AI Agents groups skills by the core loop they support:
+
+| Core Loop | What It Covers |
+|---|---|
+| Planning | Discovery, PRD, roadmap, task breakdown |
+| Architecture | Clean code, module boundaries, model separation |
+| Development | Feature implementation, UI build, refactor, code review |
+| Performance | Startup, ANR, memory, frame, battery, network, app size |
+| Security | OWASP, secrets, auth, storage, deep links, WebView, permissions, release hardening |
+| Testing | Unit tests, UI tests, accessibility, device QA, PRD verification |
+| Release | CI/CD, signing, release notes, rollout, store checklist |
+| Growth | ASO, store listing, screenshots, monetization |
+| Maintenance | Crash triage, context memory, issue-to-agent, regression checks |
+
+See [core-loops.md](core-loops.md) for the full product model.
+
+---
+
+## Categories
+
+### Code Quality
+
+Reviews, audits, and refactoring suggestions for mobile codebases.
+
+**Platforms:** Android, iOS, Flutter, React Native  
+**Examples:** Clean Architecture review, idiomatic language checks, dependency injection audit  
+**Agents:** Android Code Reviewer, Swift Code Reviewer
+
+---
+
+### Performance & Optimization
+
+Identifying and fixing performance bottlenecks — frame drops, excessive memory, slow builds, and over-the-wire inefficiency.
+
+**Platforms:** All  
+**Examples:** Re-render audit (RN), Compose recomposition, FlatList tuning, APK size reduction, GPU overdraw  
+**Agents:** React Native Performance Optimizer
+
+---
+
+### UI & Design
+
+Generating, reviewing, and improving UI code — widgets, components, layouts, animations.
+
+**Platforms:** Flutter, Android (Compose), iOS (SwiftUI), React Native  
+**Examples:** Widget generation, layout from description, animation code, dark mode audit  
+**Agents:** Flutter Widget Generator, Unity Shader Generator
+
+---
+
+### Testing & Verification
+
+Proving the app matches requirements through tests, device QA, accessibility checks, screenshots, and PRD evidence.
+
+**Platforms:** All
+
+**Examples:** PRD verification, Mobile MCP QA, accessibility checks, restart persistence tests
+
+**Skills:** `/prd-verification`, `/mobile-mcp-qa`, `/accessibility-audit`
+
+---
+
+### Debugging & Crash Analysis
+
+Parsing and explaining crash logs, stack traces, ANRs, and error states.
+
+**Platforms:** Android, iOS  
+**Examples:** NPE root cause, symbolicated iOS crash, ANR main thread analysis, OOM tracing  
+**Agents:** Crash Log Analyzer  
+**Skills:** Crash Analysis
+
+---
+
+### DevOps & Release
+
+CI/CD, deployment pipelines, store submissions, and release management.
+
+**Platforms:** All (Cross-Platform)  
+**Examples:** GitHub Actions for Android, Fastlane iOS, release notes from git log, ASO copy  
+**Agents:** CI/CD Pipeline Generator, Release Notes Generator, Store Listing Writer
+
+---
+
+### Game Development
+
+Shader authoring, Blueprint logic, game object architecture, and game-specific performance.
+
+**Platforms:** Unity, Unreal Engine  
+**Examples:** HLSL/ShaderLab shader generation, Blueprint-to-C++ migration, tick optimization  
+**Agents:** Unity Shader Generator, Unreal Blueprint Advisor
+
+---
+
+### Documentation
+
+Auto-generating documentation from code — API docs, inline comments, README files, architecture decisions.
+
+**Platforms:** All  
+**Examples:** KDoc generation, Swift DocC comments, README from project structure  
+*(Agents coming soon — contributions welcome)*
+
+---
+
+## Tagging Convention
+
+Each agent and skill file uses the following frontmatter tags:
+
+```
+**Platform:** Android | iOS | Flutter | React Native | Unity | Unreal | Cross-Platform
+**Category:** Code Quality | Performance | UI & Design | Testing | Debugging | DevOps | Game Dev | Documentation
+**Complexity:** Low | Medium | High
+```
+
+**Complexity** guide:
+- **Low** — input is simple (plain text, short code), output is focused (one type of result)
+- **Medium** — input requires structured fields, output has multiple sections
+- **High** — input is large or requires domain expertise to fill correctly, output has complex interdependent sections

package/docs/vscode-copilot-integration.md

@@ -0,0 +1,104 @@
+# Using Mobile Dev Skill Agents with GitHub Copilot (VS Code)
+
+GitHub Copilot Chat supports custom workspace instructions via `.github/copilot-instructions.md`. When this file exists in your repo, Copilot Chat reads it automatically for every `@workspace` query.
+
+This repo includes a pre-built `copilot-instructions.md` that loads the full agent index and platform conventions into Copilot.
+
+---
+
+## Setup (One-Time)
+
+The file `.github/copilot-instructions.md` is already in this repo. No configuration needed — open the repo in VS Code with the GitHub Copilot extension installed, and it is active.
+
+To verify, open Copilot Chat and ask:
+
+> @workspace What agents are available for Android development?
+
+Copilot should list the Android Code Reviewer, Android Crash Analyzer, and Android Compose UI Reviewer.
+
+---
+
+## Workflow 1 — Code Review via Copilot Chat
+
+Open the file you want reviewed, then in Copilot Chat:
+
+```
+@workspace Review the open file using the Android Code Reviewer agent rules.
+Flag CRITICAL and WARNING issues with the line number and a concrete fix for each.
+```
+
+Or reference the agent directly:
+
+```
+#file:agents/android/code-reviewer/agent.md
+Review #file:app/src/main/java/com/example/ProfileViewModel.kt
+```
+
+---
+
+## Workflow 2 — Inline Fix with Copilot Edits
+
+1. Select a problematic code block
+2. Press Cmd+I (or Ctrl+I on Windows) to open the inline chat
+3. Type:
+
+```
+Fix the coroutine scope leak and dependency injection issues identified by the Android Code Reviewer.
+```
+
+Copilot Edits will propose a diff inline — accept or discard each change.
+
+---
+
+## Workflow 3 — Crash Analysis
+
+```
+@workspace Analyze this crash log using the Android Crash Analyzer rules and return
+the 9-section report (Summary, Root Cause, Why It Happens, Risk Level, Fix, Code,
+Edge Cases, Test Checklist, Prevention Tips):
+
+[paste crash log]
+```
+
+---
+
+## Workflow 4 — Widget / Shader Generation
+
+```
+@workspace Using the Flutter Widget Generator rules, generate a Dart widget for a
+loading skeleton screen that shows placeholder shimmer cards while content loads.
+Flutter 3.27, no external packages.
+```
+
+---
+
+## Workflow 5 — @workspace for Architecture Questions
+
+```
+@workspace Using the Clean Architecture rules for Android, does this project correctly
+separate domain, data, and presentation layers? List any violations.
+```
+
+Copilot uses `@workspace` to read your actual project files alongside the agent rules from `.github/copilot-instructions.md`.
+
+---
+
+## Limitations
+
+- Copilot Chat has a shorter context window than Claude Code or GPT-4o — for very long files, split the review into sections.
+- Copilot cannot read `.cursorrules` — it reads only `.github/copilot-instructions.md`.
+- The `#file:` syntax for referencing agent files requires Copilot Chat version 1.x or later (VS Code extension).
+- Complex structured output (the full 9-section crash report) may be truncated in inline chat — use the Copilot Chat panel instead.
+
+---
+
+## Manually Adding Agent Rules to Any Copilot Session
+
+If you want to use a specific agent without `@workspace`:
+
+1. Open the agent file (e.g., `agents/ios/swift-reviewer/agent.md`)
+2. Copy the `## System Prompt` section
+3. Paste it at the top of your Copilot Chat message
+4. Follow it with your code and the input format
+
+This works in any Copilot Chat session, even outside this repo.