@leejungkiin/awkit 1.1.6 → 1.1.9

package/README.md

@@ -6,15 +6,46 @@ AWKit là framework điều phối AI agent chuyên nghiệp. Đây là **nơi d
 
 ---
 
+## 🧠 Gstack Core Principles
+
+AWKit (Antigravity v12+) thực thi 8 nguyên tắc lõi (Gstack-inspired) nhằm đảm bảo AI hoạt động an toàn và hiệu quả:
+
+1. **Anti-sycophancy:** AI phải phản biện các hướng tiếp cận có rủi ro, không đồng ý mù quáng khi có red flags.
+2. **Search Before Building (3-layer):** Tìm kiếm NeuralMemory/codebase → Tìm standard libraries/SDKs → Chỉ code mới nếu 2 bước trên thất bại.
+3. **Completion Status Protocol:** 4 trạng thái rõ ràng: `DONE`, `DONE_WITH_CONCERNS`, `BLOCKED`, `NEEDS_CONTEXT`.
+4. **6 Decision Principles:** Complete > Shortcuts, Evidence > Assumptions, Standard > Custom, Explicit > Implicit, Test > Trust, Small > Big.
+5. **3-Strike Escalation:** Sau 3 lần thử fix bug thất bại → STOP, báo cáo context và chờ user quyết định.
+6. **Scope Freeze:** Khi debugging: KHÔNG sửa code không liên quan, KHÔNG refactor. Chỉ tập trung root cause.
+7. **Boil-the-Lake Checklist:** Checklist trước khi báo DONE: error handling, edge cases, logging, cleanup, input validation...
+8. **Safety Guardrails:** Chặn auto-run cho các lệnh nguy hiểm (`rm -rf`, `git push --force`, deploy production). Bắt buộc double-confirm.
+
+---
+
 ## 🚀 Quick Start
 
-### Cài đặt nhanh qua NPM (Khuyên dùng)
+### Cài đặt (3 bước)
+
 ```bash
+# 1. AWKit Core (CLI + workflows + skills)
 npm install -g @leejungkiin/awkit
 awkit install
+
+# 2. Symphony (Task Management & Multi-Agent Orchestration)
+npm install -g @leejungkiin/awkit-symphony
+
+# 3. NeuralMemory (AI Memory Engine — requires Python >= 3.11)
+pip install neural-memory
+nmem init
+
+# Verify
 awkit doctor
+symphony --version
+nmem --version
 ```
 
+> 💡 **Bước 2 & 3 được tự động cài khi chạy `awkit install`** nếu môi trường đầy đủ.
+> AI agent cũng tự detect và cài các dependency thiếu khi init session.
+
 
 ## 📦 Commands
 
@@ -132,6 +163,25 @@ awkit tg send "Hello from AWKit!"
 awkit tg send "Hello" --chat -100123456789 --topic 1234
 ```
 
+## 🧠 GitNexus Code Intelligence
+
+AWKit (từ bản cập nhật tích hợp mới) sử dụng **GitNexus** làm engine cung cấp Code Intelligence & Knowledge Graph cho AI Agent. Sau khi chạy `npm install` AWKit, GitNexus sẽ sẵn sàng để dùng ngay.
+
+Để kích hoạt GitNexus trên bất kỳ project nào anh/chị đang code:
+
+```bash
+# 1. Quét toàn bộ project để tạo knowledge graph (chỉ cần chạy lần đầu hoặc khi cấu trúc code thay đổi lớn)
+npx gitnexus analyze
+
+# 2. Sử dụng CLI để phân tích code hoặc explore codebase:
+gitnexus query "điểm thực thi workflow"      # Tìm các file/class liên quan theo ngữ nghĩa
+gitnexus impact "tên_hàm_cần_sửa"            # Xem mức độ ảnh hưởng (blast radius) trước khi edit code!
+gitnexus context "tên_class"                 # Góc nhìn 360 độ (Ai gọi nó? Nó gọi ai?)
+gitnexus rename "tên_cũ" "tên_mới" --dry-run # Đổi tên an toàn qua biểu đồ
+```
+
+AI agent của hệ thống cũng tự động gọi qua cổng GitNexus bằng MCP Tools để tự phân tích rủi ro trước khi commit!
+
 ## 📦 Skill Packs
 
 ```bash

package/bin/awk.js

@@ -283,7 +283,7 @@ function checkSymphony({ silent = false } = {}) {
         return true;
     } catch (_) {
         if (!silent) warn('Symphony CLI not found. Please install it manually:');
-        if (!silent) dim('  npm install -g awkit-symphony');
+        if (!silent) dim('  npm install -g @leejungkiin/awkit-symphony');
         return false;
     }
 }
@@ -1797,7 +1797,7 @@ function cmdInit(forceFlag = false) {
         
         const symReady = checkSymphony({ silent: true });
         if (!symReady) {
-            dim('Symphony CLI is not installed. Run: npm i -g awkit-symphony');
+            dim('Symphony CLI is not installed. Run: npm i -g @leejungkiin/awkit-symphony');
         }
     }
 

package/core/GEMINI.md

@@ -1,6 +1,6 @@
-# GEMINI.md — Antigravity v11.0
+# GEMINI.md — Antigravity v12.1
 
-> Rules + routing only. Gate details → skills. Updated: 2026-03-18
+> Rules + routing only. Gate details → skills. Updated: 2026-03-23
 
 ---
 
@@ -50,10 +50,19 @@ Mỗi skill tự xử lý gate logic riêng — xem SKILL.md của từng skill.
 - AI models: Gemini 2.5+ only.
 - Firebase: Firebase AI Logic SDK.
 
-### Spec-First (NEW v11.0)
-- PLANNING mode PHẢI đọc `docs/specs/` **trước** khi viết `implementation_plan.md`.
-- Mỗi task trong plan NÊN có format XML `<task>` (xem `templates/specs/task-spec-template.xml`).
-- `implementation_plan.md` PHẢI reference `TECH-SPEC.md` constraints khi relevant.
+### 5-Gate Autonomous System (NEW v12.0)
+- orchestrator PHẢI triage complexity (TRIVIAL/MODERATE/COMPLEX) trước mọi task.
+- COMPLEX tasks PHẢI qua 5 Gates tuần tự:
+  - Gate 1 (Spec): `brainstorm-agent` → BRIEF.md / spec document
+  - Gate 2 (Architecture): `spec-gate` → design doc + user approve
+  - Gate 3 (Tasks): `symphony-enforcer` → tạo Symphony tickets
+  - Gate 4 (Execution): code theo ticket, đối chiếu design doc
+  - Gate 5 (Verification): `verification-gate` + `code-review`
+- TRIVIAL tasks bypass → thẳng Gate 4.
+- MODERATE tasks → Gate 3 + 4 + 5.
+- AI tự detect gate state — user KHÔNG CẦN gọi workflow bằng tay.
+- Trong lúc code, nếu cần sửa schema khác approved design → ⛔ DỪNG, quay Gate 2.
+- Chi tiết: xem `orchestrator/SKILL.md` (triage) + `spec-gate/SKILL.md` (Gate 2).
 
 ### NeuralMemory
 - Brain = projectId. Switch trước mọi nmem call.
@@ -67,6 +76,34 @@ Mỗi skill tự xử lý gate logic riêng — xem SKILL.md của từng skill.
 - Kết thúc task: Tóm tắt + Test + Next steps.
 - Không rõ: Hỏi lại, tối đa 2 lần.
 
+### Anti-sycophancy (Trung thực)
+- PHẢI push back khi approach của user có vấn đề — giải thích rõ trade-offs.
+- KHÔNG nói "Great idea!" nếu idea có red flags → nêu rủi ro trước.
+- Nêu cả ưu VÀ nhược điểm của mọi approach, không chỉ list ưu điểm.
+- Nếu request vi phạm best practices → cảnh báo TRƯỚC khi implement.
+- Khi phát hiện pattern sai → đề xuất alternative, không im lặng đồng ý.
+- Acknowledge limitations: nói "Tôi không chắc" khi không chắc.
+
+### Safety Guardrails (Destructive Commands)
+- KHÔNG BAO GIỜ set `SafeToAutoRun=true` cho các commands sau:
+  - `rm -rf`, `rm -r`, `rmdir` (recursive delete)
+  - `git push --force`, `git reset --hard`, `git clean -fd`
+  - `DROP TABLE`, `DROP DATABASE`, `DELETE FROM` (without WHERE)
+  - `docker system prune`, `docker volume rm`
+  - `npm publish`, `pod trunk push`
+  - Bất kỳ command nào deploy lên production
+- Double-confirm với user trước khi chạy destructive command.
+- Nếu không chắc command có destructive hay không → hỏi trước.
+
+### 6 Decision Principles (Auto-decide)
+Khi AI cần tự quyết định mà không hỏi user:
+1. **Complete > Shortcuts** — Implement đủ, không bỏ edge cases.
+2. **Evidence > Assumptions** — Dựa trên data, không đoán.
+3. **Standard > Custom** — Ưu tiên solution có sẵn, chuẩn ngành.
+4. **Explicit > Implicit** — Code rõ ràng, không "clever" tricks.
+5. **Test > Trust** — Viết test, không "chắc chắn đúng".
+6. **Small > Big** — Incremental changes, không big-bang refactor.
+
 ### Project Context
 - CODEBASE.md tồn tại → KHÔNG scan raw directory.
 - KHÔNG hỏi user về project structure.
@@ -85,7 +122,8 @@ Mỗi skill tự xử lý gate logic riêng — xem SKILL.md của từng skill.
 
 ## Routing
 
-- **Execution order:** `symphony-orchestrator` → `awf-session-restore` → `nm-memory-sync` → `symphony-enforcer` → `orchestrator` → action
+- **Execution order:** `symphony-orchestrator` → `awf-session-restore` → `nm-memory-sync` → `symphony-enforcer` → `orchestrator` (triage + gate-check) → action
+- **Gate skills:** `orchestrator` (triage) → `brainstorm-agent` (G1) → `spec-gate` (G2) → `symphony-enforcer` (G3) → `verification-gate` (G5)
 - **Skill catalog:** xem `orchestrator/SKILL.md`
 - **Workflows:** 75+ (`/xxx`). Core: `/init` `/code` `/debug` `/recap` `/next` `/todo`
 - **Shortcuts:** `/todo` `/done` `/next`

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@leejungkiin/awkit",
-  "version": "1.1.6",
+  "version": "1.1.9",
   "description": "Antigravity Workflow Kit. Unified AI agent orchestration system.",
   "main": "bin/awk.js",
   "bin": {
@@ -19,7 +19,7 @@
   "keywords": [
     "awkit"
   ],
-  "author": "Kien AI",
+  "author": "Kien Le",
   "license": "MIT",
   "engines": {
     "node": ">=18.0.0"
@@ -33,9 +33,12 @@
     "schemas/",
     "templates/",
     "docs/",
-    "symphony/",
     "README.md",
     "CHANGELOG.md",
     "VERSION"
-  ]
-}
+  ],
+  "dependencies": {
+    "@leejungkiin/awkit-symphony": "^0.1.0",
+    "gitnexus": "^1.4.7"
+  }
+}

package/skill-packs/neural-memory/skills/nm-memory-sync/SKILL.md

@@ -165,9 +165,22 @@ This surfaces memories that a keyword search would miss.
 
 ---
 
+## Auto-Install Protocol (New Machine)
+
+If `nmem` CLI is not available:
+```
+1. AI detects: command not found: nmem
+2. AI auto-runs: pip install neural-memory (or python3 -m pip install neural-memory)
+   → Requires Python >= 3.11
+3. After install: nmem init (initialize brain DB)
+4. Verify: nmem --version
+5. If Python < 3.11 or pip fails:
+   → Show user: "⚠️ NeuralMemory requires Python >= 3.11. Install via: brew install python@3.12"
+```
+
 ## Fallback Mode (NeuralMemory Not Installed)
 
-If `nmem` is not available, falls back to flat-file behavior:
+If `nmem` is not available AND auto-install fails, falls back to flat-file behavior:
 ```
 - Reads from brain/decisions/, brain/solutions/
 - Keyword overlap matching (legacy behavior)

package/skills/ab-test-store-listing/SKILL.md

@@ -0,0 +1,220 @@
+---
+name: ab-test-store-listing
+description: When the user wants to A/B test App Store product page elements to improve conversion rate. Also use when the user mentions "A/B test", "product page optimization", "test my screenshots", "test my icon", "conversion rate optimization", "CPP", or "custom product pages". For screenshot design, see screenshot-optimization. For metadata optimization, see metadata-optimization.
+metadata:
+  version: 1.0.0
+---
+
+# A/B Test Store Listing
+
+You are an expert in App Store product page optimization and A/B testing. Your goal is to help the user design, run, and interpret tests that improve their App Store conversion rate.
+
+## Initial Assessment
+
+1. Check for `app-marketing-context.md` — read it for context
+2. Ask for the **App ID**
+3. Ask for **current conversion rate** (if known from App Store Connect)
+4. Ask for **daily impressions** (determines test duration)
+5. Ask: **What do you want to test?** (icon, screenshots, description, etc.)
+
+## What You Can Test
+
+### Apple Product Page Optimization (PPO)
+
+Apple's native A/B testing tool in App Store Connect.
+
+| Element | Testable? | Notes |
+|---------|-----------|-------|
+| App icon | Yes | Up to 3 variants |
+| Screenshots | Yes | Up to 3 variants |
+| App preview video | Yes | Up to 3 variants |
+| Description | No | Not testable via PPO |
+| Title | No | Not testable via PPO |
+| Subtitle | No | Not testable via PPO |
+
+**Limitations:**
+- Only tests against organic App Store traffic
+- Minimum 90% confidence required to declare winner
+- Tests run for 7-90 days
+- Can only run one test at a time
+- Traffic split is automatic (not configurable)
+
+### Custom Product Pages (CPP)
+
+35 custom product pages per app, each with unique:
+- Screenshots
+- App preview videos
+- Promotional text
+
+**Use for:**
+- Different audiences (from different ad campaigns)
+- Different value propositions
+- Seasonal messaging
+- Localized creative for specific markets
+
+**Not a true A/B test** — CPPs are targeted pages linked from specific URLs/campaigns, not random traffic splits.
+
+## Test Prioritization
+
+### Impact × Effort Matrix
+
+| Element | Impact on CVR | Effort | Priority |
+|---------|--------------|--------|----------|
+| First screenshot | Very High (15-30% lift possible) | Medium | 1 |
+| App icon | High (10-20% lift possible) | Medium | 2 |
+| Screenshot order | Medium (5-15% lift possible) | Low | 3 |
+| Screenshot style | Medium (5-15% lift possible) | High | 4 |
+| Preview video | Medium (5-10% lift possible) | High | 5 |
+
+### What to Test First
+
+**Always start with the first screenshot.** It has the highest impact because:
+- It's the first thing users see in search results
+- 80% of users never scroll past the first 3 screenshots
+- Small improvements here affect every visitor
+
+## Test Design Framework
+
+### Step 1: Hypothesis
+
+Write a clear hypothesis before each test:
+
+```
+If we [change], then [metric] will [improve/increase] because [reason].
+```
+
+**Examples:**
+- "If we add social proof ('5M+ users') to the first screenshot, conversion rate will increase because it builds trust"
+- "If we change the icon from blue to orange, tap-through rate will increase because it stands out more in search results"
+- "If we show the app's AI feature first instead of the basic editor, conversion will increase because AI is the key differentiator"
+
+### Step 2: Variants
+
+Design 2-3 variants (including control):
+
+| Variant | Description | Hypothesis |
+|---------|-------------|------------|
+| Control (A) | Current version | Baseline |
+| Variant B | [specific change] | [why it might win] |
+| Variant C | [different change] | [why it might win] |
+
+**Rules for good variants:**
+- Change ONE thing per test (isolate the variable)
+- Make the change significant enough to detect (don't test subtle color shifts)
+- Each variant should have a clear hypothesis
+- Don't test more than 3 variants (dilutes traffic)
+
+### Step 3: Sample Size
+
+Calculate required test duration:
+
+```
+Daily impressions: [N]
+Current conversion rate: [X]%
+Minimum detectable effect: [Y]% (relative improvement)
+Confidence level: 95%
+
+Required sample per variant: ~[N] impressions
+Estimated duration: [N] days
+```
+
+**Rules of thumb:**
+- < 1000 daily impressions: Tests take 30-90 days (consider if worth it)
+- 1000-5000 daily impressions: Tests take 14-30 days
+- 5000+ daily impressions: Tests take 7-14 days
+- Need at least 1000 impressions per variant for meaningful results
+
+### Step 4: Run the Test
+
+**In App Store Connect:**
+1. Go to Product Page Optimization
+2. Create a new test
+3. Upload variant assets
+4. Set test duration (recommend: let it run until statistical significance)
+5. Monitor but don't stop early
+
+### Step 5: Interpret Results
+
+**Statistical significance:**
+- Apple requires 90% confidence minimum
+- Aim for 95% confidence before making decisions
+- Look at the confidence interval, not just the point estimate
+
+**What to look for:**
+- Conversion rate lift (primary metric)
+- Impression-to-tap rate (for icon tests)
+- Download rate (for screenshot/video tests)
+- Segment differences (new vs returning, country, source)
+
+## Common Test Ideas
+
+### Icon Tests
+
+| Test | Control | Variant | Expected Impact |
+|------|---------|---------|----------------|
+| Color | Current color | Contrasting color | 5-20% TTR change |
+| Style | Detailed | Simplified | 5-15% TTR change |
+| Element | Current symbol | Different symbol | 5-20% TTR change |
+| Background | Solid | Gradient | 3-10% TTR change |
+
+### Screenshot Tests
+
+| Test | Control | Variant | Expected Impact |
+|------|---------|---------|----------------|
+| First screenshot | Feature-focused | Benefit-focused | 10-30% CVR change |
+| Social proof | No social proof | "5M+ users" badge | 5-15% CVR change |
+| Text size | Small text | Large, bold text | 5-10% CVR change |
+| Style | Light mode | Dark mode | 5-15% CVR change |
+| Layout | Device frame | Full-bleed | 5-10% CVR change |
+| Order | Current order | Reordered by benefit | 5-15% CVR change |
+
+### Video Tests
+
+| Test | Control | Variant | Expected Impact |
+|------|---------|---------|----------------|
+| Has video | No video | 15s feature demo | 5-15% CVR change |
+| Hook | Feature demo | Problem/solution | 5-10% CVR change |
+| Length | 30s | 15s | 3-8% CVR change |
+
+## Output Format
+
+### Test Plan
+
+```
+Test Name: [descriptive name]
+Element: [icon / screenshots / video]
+Hypothesis: If we [change], then [metric] will [improve] because [reason]
+
+Variants:
+- Control (A): [description]
+- Variant B: [description]
+- Variant C: [description] (optional)
+
+Estimated Duration: [N] days
+Required Impressions: [N] per variant
+Success Metric: [conversion rate / tap-through rate]
+Minimum Detectable Effect: [X]%
+```
+
+### Test Results Interpretation
+
+When the user shares results:
+1. Is it statistically significant? (confidence level)
+2. What's the actual lift? (with confidence interval)
+3. Are there segment differences?
+4. What's the next test to run?
+5. Estimated annual impact (downloads × lift)
+
+### Testing Roadmap
+
+Provide a 3-month testing calendar:
+- Month 1: [highest impact test]
+- Month 2: [second priority test]
+- Month 3: [third priority test]
+
+## Related Skills
+
+- `screenshot-optimization` — Design screenshot variants
+- `metadata-optimization` — Optimize non-testable elements
+- `app-analytics` — Track conversion metrics
+- `aso-audit` — Identify what to test first

package/skills/android-aso/SKILL.md

@@ -0,0 +1,197 @@
+---
+name: android-aso
+description: When the user wants to optimize their Google Play Store listing — title, short description, full description, keywords, ratings, or Play Store-specific features. Use when the user mentions "Google Play", "Android", "Play Store", "Play Console", "short description", "full description indexed", "Google Play ASO", or wants Google Play-specific keyword, creative, or ratings strategy. For iOS App Store optimization, see aso-audit and metadata-optimization.
+metadata:
+  version: 1.0.0
+---
+
+# Android ASO (Google Play)
+
+You are a Google Play ASO expert. Google Play's algorithm differs fundamentally from iOS — the full description is indexed, there is no hidden keyword field, and ratings are continuous (not version-reset).
+
+## Key Differences vs iOS
+
+| Factor | Google Play | Apple App Store |
+|--------|------------|----------------|
+| **Keyword indexing** | Title + Short desc + Full desc (all indexed) | Title + Subtitle + Keyword field only |
+| **Hidden keyword field** | ✗ None | ✓ 100-char field |
+| **Description indexed** | ✓ Full 4000 chars | ✗ Not indexed |
+| **Ratings** | Continuous — never reset | Reset per version (can request reset) |
+| **A/B testing** | Play Store Experiments (native) | Product Page Optimization |
+| **Screenshots** | 2–8 per language | Up to 10 per language |
+| **Feature graphic** | Required (1024×500px) | Not applicable |
+| **Algorithm signals** | Installs, engagement, ratings, keywords | Keyword match, ratings, conversions |
+| **Review indexing** | Reviews and replies indexed | Not indexed |
+
+## Character Limits
+
+| Field | Limit | Indexed | Weight |
+|-------|-------|---------|--------|
+| **Title** | 30 chars | ✓ | Highest |
+| **Short description** | 80 chars | ✓ | High |
+| **Full description** | 4000 chars | ✓ | Medium |
+| **Developer name** | — | ✓ | Low |
+
+## Initial Assessment
+
+1. Check for `app-marketing-context.md`
+2. Ask: **Do you have Play Console access?** (for actual keyword data)
+3. Ask: **What is your current title and short description?**
+4. Ask: **What are your 3 most important keywords?**
+5. Ask: **What category is your app in?**
+
+## Metadata Optimization
+
+### Title (30 chars)
+
+- Lead with your brand name or primary keyword — whichever is stronger
+- Include 1 high-volume keyword naturally: `Brand – Keyword Descriptor`
+- ✅ "Headspace: Meditation & Sleep" | ❌ "Best Meditation App for You"
+
+### Short Description (80 chars)
+
+- First thing users read on search results
+- Pack your 2–3 most important keywords naturally in one compelling sentence
+- ✅ "Guided meditation, sleep sounds & breathing exercises for stress relief"
+- Do not repeat the title's primary keyword
+
+### Full Description (4000 chars — indexed)
+
+**Structure for algorithm + conversion:**
+
+```
+[Hook paragraph — 2–3 sentences]
+Lead with the core value proposition. Include primary keyword in first 167 chars
+(shown above the fold).
+
+[Feature bullets — 5–8 items]
+• [Feature]: [Benefit]
+Use keywords naturally. Vary phrasing — don't repeat exact phrases.
+
+[Social proof]
+"Trusted by X million users" / awards / press mentions
+
+[Call to action]
+Download [App Name] today — [value prop].
+
+[Keywords section — natural, not stuffed]
+A paragraph using keyword variants, synonyms, and long-tail terms.
+```
+
+**Keyword density rule:** Target keyword should appear 3–5 times across the full description. Exact match + variants. Never stuff.
+
+### Localization
+
+Google Play indexes descriptions per language. Each locale is a fresh keyword opportunity — translate and localize, don't just auto-translate.
+
+## Keyword Research for Play Store
+
+Use Appeeky keyword tools, then adapt for Play:
+
+```bash
+GET /v1/keywords/metrics?keywords=meditation,mindfulness,sleep sounds&country=us
+GET /v1/keywords/suggestions?term=meditation&country=us
+```
+
+**Play-specific considerations:**
+- Long-tail phrases work well (full description is indexed)
+- Semantic similarity matters — Google's algorithm understands synonyms
+- User reviews and Q&A also get indexed — common words in reviews can signal keywords
+
+## Feature Graphic (1024×500px)
+
+Required for Play Store. Appears at the top of your listing when no video is present.
+
+- Show the core use case in one image
+- Text is legible — no tiny copy
+- Brand-consistent with screenshots
+- Works without text (text may be truncated on some surfaces)
+
+## Ratings Strategy
+
+Unlike iOS, Play ratings are **never reset** — every rating ever given counts.
+
+**To improve your rating:**
+1. Respond to every 1–3 star review (boosts score algorithmically)
+2. Reply invites re-rating — users can update their review
+3. Use `review-management` skill for response templates
+4. Fix the issues mentioned in low ratings and reply: "Fixed in version X.X"
+
+**Rating prompt timing** (see also `rating-prompt-strategy` skill):
+- Prompt after a clear success moment, not on cold open
+- Use the Play In-App Review API: `ReviewManager.requestReviewFlow()`
+
+## Play Store Experiments (A/B Testing)
+
+Native A/B testing for:
+- Icon
+- Feature graphic
+- Screenshots (up to 3 variants)
+- Short description (up to 3 variants)
+- Full description (up to 3 variants)
+
+Access: Play Console → Store listing experiments
+
+**Test one element at a time. Run for minimum 7 days or 1,000 impressions.**
+
+## Pre-Launch (Early Access)
+
+Use Early Access to:
+- Collect reviews before public launch
+- Get indexed by Google before launch
+- Get editorial consideration from Google Play
+
+## Output Format
+
+### Play Store Listing Draft
+
+```
+Title (30):     [text]
+Short desc (80): [text]
+
+Full Description:
+[Hook — 2–3 sentences, primary keyword in first 167 chars]
+
+✨ Features:
+• [Feature]: [Benefit]
+• [Feature]: [Benefit]
+• [Feature]: [Benefit]
+• [Feature]: [Benefit]
+• [Feature]: [Benefit]
+
+[Social proof paragraph]
+
+[CTA sentence]
+
+[Keyword-rich closing paragraph]
+
+Keywords targeted: [list primary keywords used]
+```
+
+### ASO Audit (Play)
+
+Score each field 1–10:
+
+```
+Title:             [N]/10 — [note]
+Short description: [N]/10 — [note]
+Full description:  [N]/10 — [note]
+Screenshots:       [N]/10 — [note]
+Feature graphic:   [N]/10 — [note]
+Ratings:           [N]/10 — [note]
+Overall:           [N]/60
+
+Top 3 improvements:
+1. [specific change with expected impact]
+2. [specific change with expected impact]
+3. [specific change with expected impact]
+```
+
+## Related Skills
+
+- `aso-audit` — iOS-focused audit (compare approaches)
+- `metadata-optimization` — iOS metadata (different field rules)
+- `review-management` — Respond to Play reviews to recover rating
+- `rating-prompt-strategy` — In-App Review API timing and strategy
+- `ab-test-store-listing` — Play Experiments methodology
+- `localization` — Per-language listing optimization