mobile-ai-agents 1.0.23

package/skills/cross-platform/mobile-memory-save.md

@@ -0,0 +1,210 @@
+# Skill — /mobile-memory-save
+
+**Platform:** Cross-Platform  
+**Slash Command:** `/mobile-memory-save`  
+**Composable With:** Mobile Memory, AXIOM, CRASHER, SWIFT, DART, BRIDGE, PERF
+
+---
+
+## Purpose
+
+Create a compact `MOBILE_MEMORY.md` checkpoint fast. Use this when the session is getting long, a major decision was made, or you need to switch AI tools without losing progress.
+
+This skill does not require activating the full Mobile Memory agent. It produces a smaller file with only:
+
+- `INSTANT RESUME`
+- `Progress`
+- `NEXT ACTION`
+- `Code State`
+- Resume instructions
+
+---
+
+## Input Format
+
+```text
+COMMAND: /mobile-memory-save
+PROJECT: <project name>
+PLATFORM: <Android/iOS/Flutter/RN/Unity/Unreal>
+CURRENT TASK:
+<what we are doing>
+DECISIONS:
+<bullet list of decisions made>
+PROGRESS:
+<done / in progress / blocked>
+CODE STATE:
+<files changed, files mid-edit, important snippets if needed>
+NEXT ACTION:
+<one executable instruction>
+```
+
+If the user does not provide fields, infer them from the conversation and mark uncertain items as `AMBIGUOUS`.
+
+---
+
+## Skill Prompt
+
+```text
+Create a compact MOBILE_MEMORY.md checkpoint from the current session.
+
+Rules:
+1. Capture only what a future AI needs to continue without asking the user to repeat context.
+2. Do not summarize away decisions, blockers, or file names.
+3. Produce exactly one NEXT ACTION. It must be executable, concrete, and free of assumed context.
+4. Include code state for any file mid-edit or any snippet that would be lost between sessions.
+5. Use the compact MOBILE_MEMORY.md format below. Do not add extra sections.
+
+Compact MOBILE_MEMORY.md format:
+
+---
+# Mobile Memory
+**Project:** [name]
+**Platform:** [Android/iOS/Flutter/RN/Unity/Unreal]
+**Stack:** [language, framework, key libraries]
+**Architecture:** [pattern]
+**Saved:** [date]
+**Compatible:** Claude Code · Cursor · Windsurf · ChatGPT · Gemini · Copilot
+**Token reduction:** [X× vs reading raw files]
+
+---
+
+## ⚡ INSTANT RESUME
+[2-4 self-contained sentences covering project, task, decisions, progress, and next move.]
+
+---
+
+## 🎯 Session Context
+
+### Current Task
+[Exact task in one paragraph]
+
+### Decisions Made
+| Decision | Reason | Rejected |
+|---|---|---|
+
+### Progress
+✅ Done: [list]
+🔄 In Progress: [exact state of anything mid-edit]
+⏭️ NEXT ACTION: [specific, executable, no assumed context]
+🚧 Blocked: [list or "Nothing"]
+
+### Open Questions
+- [question 1]
+- [question 2]
+
+---
+
+## 📄 Code State
+[For any file mid-edit, paste current state here]
+
+---
+
+## 🔄 Resume Instructions
+
+**Claude Code:**
+Start new session → paste INSTANT RESUME → paste full MOBILE_MEMORY.md → say "Continue"
+
+**Cursor/Windsurf:**
+Save as MOBILE_MEMORY.md in project root → next prompt: "Read MOBILE_MEMORY.md and continue"
+
+**ChatGPT/Gemini:**
+Paste full file as first message → "Resume from NEXT ACTION"
+
+**Same tool, new session:**
+Paste full file → /mobile-memory restore
+---
+```
+
+---
+
+## Output Format
+
+Return a single Markdown document named `MOBILE_MEMORY.md` using the compact format in the prompt.
+
+---
+
+## Example
+
+### Input
+
+```text
+COMMAND: /mobile-memory-save
+PROJECT: PayTrack
+PLATFORM: Flutter
+CURRENT TASK: Replace setState checkout form with Riverpod StateNotifier.
+DECISIONS:
+- Keep checkout validation in CheckoutController.
+- Use AsyncValue for submit state.
+- Keep navigation in the widget after successful submit.
+PROGRESS:
+- Done: controller skeleton created.
+- In progress: widget still calls setState in _submit().
+- Blocked: none.
+CODE STATE:
+lib/checkout/checkout_page.dart has old _submitting bool.
+NEXT ACTION: Replace _submitting bool with ref.watch(checkoutControllerProvider).isLoading and call controller.submit(orderId).
+```
+
+### Output
+
+```markdown
+---
+# Mobile Memory
+**Project:** PayTrack
+**Platform:** Flutter
+**Stack:** Dart, Flutter, Riverpod
+**Architecture:** Feature-first MVVM with StateNotifier controllers
+**Saved:** 2026-06-06
+**Compatible:** Claude Code · Cursor · Windsurf · ChatGPT · Gemini · Copilot
+**Token reduction:** 18× vs reading raw files
+
+---
+
+## ⚡ INSTANT RESUME
+PayTrack is a Flutter app refactoring checkout from local setState to a Riverpod StateNotifier. The session decided validation belongs in CheckoutController, submit state should use AsyncValue, and navigation remains in the widget after success. The controller skeleton exists, but checkout_page.dart still uses a local _submitting bool. Continue by replacing _submitting with provider state and routing submit through CheckoutController.submit(orderId).
+
+---
+
+## 🎯 Session Context
+
+### Current Task
+Replace the checkout page's local setState submit flow with a Riverpod StateNotifier while preserving existing navigation behavior.
+
+### Decisions Made
+| Decision | Reason | Rejected |
+|---|---|---|
+| Keep validation in CheckoutController | Makes submit logic testable | Validation inside widget callbacks |
+| Use AsyncValue for submit state | Represents loading, success, and error cleanly | Separate bool and error fields |
+| Keep navigation in widget | UI owns navigation after controller success | Controller using BuildContext |
+
+### Progress
+✅ Done: CheckoutController skeleton created.
+🔄 In Progress: checkout_page.dart still owns _submitting and _submit().
+⏭️ NEXT ACTION: Replace _submitting bool in lib/checkout/checkout_page.dart with ref.watch(checkoutControllerProvider).isLoading and call ref.read(checkoutControllerProvider.notifier).submit(orderId) from the submit button.
+🚧 Blocked: Nothing
+
+### Open Questions
+- Should submit errors render inline or as a SnackBar?
+
+---
+
+## 📄 Code State
+lib/checkout/checkout_page.dart still has a private `_submitting` bool and `_submit()` method that call setState before and after the API call.
+
+---
+
+## 🔄 Resume Instructions
+
+**Claude Code:**
+Start new session → paste INSTANT RESUME → paste full MOBILE_MEMORY.md → say "Continue"
+
+**Cursor/Windsurf:**
+Save as MOBILE_MEMORY.md in project root → next prompt: "Read MOBILE_MEMORY.md and continue"
+
+**ChatGPT/Gemini:**
+Paste full file as first message → "Resume from NEXT ACTION"
+
+**Same tool, new session:**
+Paste full file → /mobile-memory restore
+---
+```

package/skills/cross-platform/mobile-memory-search.md

@@ -0,0 +1,154 @@
+# Skill — /mobile-memory-search
+
+**Platform:** Cross-Platform  
+**Slash Command:** `/mobile-memory-search`  
+**Composable With:** Mobile Memory, MOBILE-HARNESS, APPFORGE, `/mobile-memory-save`, `/mobile-memory-graph`
+
+---
+
+## Purpose
+
+Search and inject local Mobile AI Agents memory captured in `.mobile-ai-agents/memory/` so an AI session can resume with the right project context without reading every file or asking the user to repeat decisions.
+
+Use this when a project has been worked on across multiple days, tools, agents, or tasks.
+
+---
+
+## Input Format
+
+```text
+COMMAND: /mobile-memory-search
+QUERY:
+<decision, feature, bug, file name, agent finding, release step, or task>
+MEMORY_SOURCE:
+<optional output from `mobile-ai-agents memory search`, `timeline`, or `inject`>
+NEXT_STEP:
+<optional current task>
+```
+
+---
+
+## Skill Prompt
+
+```text
+Use local Mobile AI Agents memory to restore relevant context.
+
+If the user has not provided memory output, ask them to run one of:
+
+- `npx mobile-ai-agents memory inject`
+- `npx mobile-ai-agents memory search <query>`
+- `npx mobile-ai-agents memory timeline --limit 20`
+
+When memory output is provided:
+
+1. Identify the project, platform, active feature, and most recent next action.
+2. Pull out decisions, rejected alternatives, unresolved findings, files touched, and blockers.
+3. Separate confirmed memory from assumptions.
+4. Recommend exact files to read before editing.
+5. Continue from the latest NEXT ACTION if it is clear.
+
+Output exactly:
+
+MOBILE MEMORY SEARCH RESULT
+=====================
+Query:
+Project:
+Platform:
+
+Relevant Memory:
+| Time | Type | Title | Why It Matters |
+|---|---|---|---|
+
+Decisions:
+- <decision or "None found">
+
+Open Risks:
+- <risk or "None found">
+
+Files To Read Next:
+- <file or "Unknown">
+
+Confirmed Context:
+- <fact>
+
+Assumptions:
+- <assumption or "None">
+
+NEXT ACTION:
+1. <single executable action>
+```
+
+---
+
+## CLI Workflow
+
+```bash
+npx mobile-ai-agents memory init
+npx mobile-ai-agents memory capture --type decision --title "Use Room" --text "Persist habits locally with Room."
+npx mobile-ai-agents memory capture --type next-action --text "Implement HabitDao and restart persistence test."
+npx mobile-ai-agents memory search persistence
+npx mobile-ai-agents memory inject
+npx mobile-ai-agents memory checkpoint
+```
+
+---
+
+## Example
+
+### Input
+
+```text
+COMMAND: /mobile-memory-search
+QUERY: persistence
+MEMORY_SOURCE:
+Project: Habit Pulse
+Platform: Android
+Events: 3
+
+## Recent Memory
+- 2026-07-09T04:00:00Z · decision · Use Room — Persist habits locally with Room instead of SharedPreferences.
+- 2026-07-09T04:03:00Z · finding · PRD verification failed — Restart persistence is missing.
+- 2026-07-09T04:05:00Z · next-action · Implement HabitDao — Implement HabitDao and restart persistence test.
+```
+
+### Output
+
+```text
+MOBILE MEMORY SEARCH RESULT
+=====================
+Query: persistence
+Project: Habit Pulse
+Platform: Android
+
+Relevant Memory:
+| Time | Type | Title | Why It Matters |
+|---|---|---|---|
+| 2026-07-09T04:00:00Z | decision | Use Room | Persistence approach chosen |
+| 2026-07-09T04:03:00Z | finding | PRD verification failed | Restart behavior missing |
+| 2026-07-09T04:05:00Z | next-action | Implement HabitDao | Current task |
+
+Decisions:
+- Use Room instead of SharedPreferences for habit persistence.
+
+Open Risks:
+- Restart persistence has not passed PRD verification.
+
+Files To Read Next:
+- Habit entity, DAO, database, repository, and current home screen state files.
+
+Confirmed Context:
+- The next implementation step is HabitDao plus restart persistence test.
+
+Assumptions:
+- Exact file paths were not present in memory output.
+
+NEXT ACTION:
+1. Read persistence-related files, then implement HabitDao and a restart persistence test.
+```
+
+---
+
+## Notes
+
+- This skill searches local memory output; it does not replace reading source files before editing.
+- Keep secrets out of memory captures. Wrap sensitive text in `<private>...</private>` before capture.

package/skills/cross-platform/perf-audit.md

@@ -0,0 +1,61 @@
+# Skill — /perf-audit
+
+**Platform:** Cross-Platform
+**Slash Command:** `/perf-audit`
+**Composable With:** agents/cross-platform/perf/agent.md
+
+---
+
+## Purpose
+
+Systematic slow-screen profiling checklist for any mobile platform. Run this before guessing at optimizations.
+
+---
+
+## Skill Prompt
+
+```
+Run a systematic performance audit on the described slow screen:
+
+STEP 1 — BASELINE MEASUREMENT (before touching code)
+Tell me the current measurements (or ask the developer to measure):
+- Android: record with Android GPU Inspector or Systrace. Look for frames > 16ms.
+- iOS: Instruments → Time Profiler. Look for main thread > 8ms between frames.
+- Flutter: DevTools → Performance tab. Flag any frame > 16ms in the flame chart.
+- RN: Flipper → Perf Monitor or React DevTools Profiler.
+Without a baseline, we cannot know if our fix worked.
+
+STEP 2 — MAIN THREAD ANALYSIS
+Is the main/UI thread being blocked?
+Check for:
+- Synchronous disk reads (SharedPreferences.getX(), NSUserDefaults, MMKV)
+- JSON parsing of large payloads on the UI thread
+- Image decoding without background threading
+- Database queries (Room, CoreData, SQLite) without async/await
+- Any operation > 1ms on the thread that drives the UI
+
+STEP 3 — RENDERING ANALYSIS (platform-specific)
+Android/Compose: Layout Inspector → check for overdraw, deep view trees.
+iOS/SwiftUI: Instruments → Core Animation. Look for offscreen renders.
+Flutter: DevTools → Widget Inspector. Check rebuild count per frame.
+React Native: Check JS thread usage in Flipper. Flag non-native animations.
+
+STEP 4 — MEMORY ANALYSIS
+Is GC/ARC pressure causing jank?
+- Profile allocation rate during the slow interaction.
+- Identify allocations per frame in hot path.
+- Identify retained objects (memory leaks) causing eventual OOM.
+
+STEP 5 — LIST PERFORMANCE (if applicable)
+- Is a non-virtualized list rendering all items?
+- Are list items creating expensive objects (network requests, DB calls) per cell?
+- Is the list re-rendering all cells on parent state change?
+
+STEP 6 — PRIORITIZED FIX LIST
+For each bottleneck found:
+1. What is it? (be specific — function name, line)
+2. Estimated time saved (ms per frame)
+3. Concrete fix with code
+
+Output measurements first, then fixes.
+```