vibes-plug 1.0.0

package/skills/seo/references/schema-types.md

@@ -0,0 +1,118 @@
+<!-- Updated: 2026-02-07 -->
+# Schema.org Types: Status & Recommendations (February 2026)
+
+**Schema.org Version:** 29.4 (December 8, 2025)
+
+## Format Preference
+Always use **JSON-LD** (`<script type="application/ld+json">`).
+Google's documentation explicitly recommends JSON-LD over Microdata and RDFa.
+
+**AI Search Note:** Content with proper schema has ~2.5× higher chance of appearing in AI-generated answers (confirmed by Google and Microsoft, March 2025).
+
+---
+
+## Active: Recommend freely
+
+| Type | Use Case | Key Properties |
+|------|----------|----------------|
+| Organization | Company info | name, url, logo, contactPoint, sameAs |
+| LocalBusiness | Physical businesses | name, address, telephone, openingHours, geo, priceRange |
+| SoftwareApplication | Desktop/mobile apps | name, operatingSystem, applicationCategory, offers, aggregateRating |
+| WebApplication | Browser-based SaaS | name, applicationCategory, offers, browserRequirements, featureList |
+| Product | Physical/digital products | name, image, description, sku, brand, offers, review |
+| Offer | Pricing | price, priceCurrency, availability, url, validFrom |
+| Service | Service businesses | name, provider, areaServed, description, offers |
+| Article | Blog posts, news | headline, author, datePublished, dateModified, image, publisher |
+| BlogPosting | Blog content | Same as Article + blog-specific context |
+| NewsArticle | News content | Same as Article + news-specific context |
+| Review | Individual reviews | reviewRating, author, itemReviewed, reviewBody |
+| AggregateRating | Rating summaries | ratingValue, reviewCount, bestRating, worstRating |
+| BreadcrumbList | Navigation | itemListElement with position, name, item |
+| WebSite | Site-level | name, url, potentialAction (SearchAction for sitelinks search) |
+| WebPage | Page-level | name, description, datePublished, dateModified |
+| Person | Author/team | name, jobTitle, url, sameAs, image, worksFor |
+| ContactPage | Contact pages | name, url |
+| VideoObject | Video content | name, description, thumbnailUrl, uploadDate, duration, contentUrl |
+| ImageObject | Image content | contentUrl, caption, creator, copyrightHolder |
+| Event | Events | name, startDate, endDate, location, organizer, offers |
+| JobPosting | Job listings | title, description, datePosted, hiringOrganization, jobLocation |
+| Course | Educational content | name, description, provider, hasCourseInstance |
+| DiscussionForumPosting | Forum threads | headline, author, datePublished, text, url |
+| ProductGroup | Variant products | name, productGroupID, variesBy, hasVariant |
+| ProfilePage | Author/creator profiles | mainEntity (Person), name, url, description, sameAs |
+
+---
+
+## Restricted: Only for specific site types
+
+| Type | Restriction | Since |
+|------|------------|-------|
+| FAQPage | Government and healthcare authority sites ONLY | August 2023 |
+
+> Google severely limited FAQ rich results in August 2023. Only authoritative sources (government, health organizations) receive FAQ rich results.
+>
+> **GEO nuance**: FAQPage schema still benefits AI/LLM citation visibility (ChatGPT, Perplexity, Google AI Overviews), even without Google rich results.
+> - **Existing FAQPage on commercial site**: Flag at Info priority, not Critical. Removal removes GEO citation upside.
+> - **Adding new FAQPage**: Not recommended for Google benefit; acceptable if AI search visibility is a priority.
+
+---
+
+## Deprecated: Never recommend
+
+| Type | Status | Since | Notes |
+|------|--------|-------|-------|
+| HowTo | Rich results fully removed | September 2023 | Google stopped showing how-to rich results |
+| SpecialAnnouncement | Deprecated | July 31, 2025 | COVID-era schema, no longer processed |
+| CourseInfo | Retired from rich results | June 2025 | Merged into Course |
+| EstimatedSalary | Retired from rich results | June 2025 | No longer displayed |
+| LearningVideo | Retired from rich results | June 2025 | Use VideoObject instead |
+| ClaimReview | Retired from rich results | June 2025 | Fact-check markup no longer generates rich results |
+| VehicleListing | Retired from rich results | June 2025 | Vehicle listing structured data discontinued |
+| Book Actions | Deprecated then REVERSED | June 2025 | **Still functional as of Feb 2026**: historical note only |
+| Practice Problem | Retired from rich results | Late 2025 | Educational practice problems no longer displayed |
+| Dataset | Retired from rich results | Late 2025 | Dataset Search feature discontinued |
+
+---
+
+## Recent Additions (2024-2026)
+
+| Type/Feature | Added | Notes |
+|-------------|-------|-------|
+| Product Certification markup | April 2025 | Energy ratings, safety certifications. Replaced EnergyConsumptionDetails. |
+| ProductGroup | 2025 | E-commerce product variants with variesBy, hasVariant properties |
+| ProfilePage | 2025 | Author/creator profile pages with mainEntity Person for E-E-A-T |
+| DiscussionForumPosting | 2024 | For forum/community content |
+| Speakable | Updated 2024 | For voice search optimization |
+| LoyaltyProgram | June 2025 | Member pricing, loyalty card structured data |
+| Organization-level shipping/return policies | November 2025 | Configure via Search Console without Merchant Center |
+| ConferenceEvent | December 2025 | Schema.org v29.4 addition |
+| PerformingArtsEvent | December 2025 | Schema.org v29.4 addition |
+
+## E-commerce Requirements (Updated)
+
+| Requirement | Status | Since |
+|-------------|--------|-------|
+| `returnPolicyCountry` in MerchantReturnPolicy | **Required** | March 2025 |
+| Product variant structured data | Expanded | 2025, includes apparel, cosmetics, electronics |
+
+> **Note:** Content API for Shopping sunsets August 18, 2026. Migrate to Merchant API.
+
+---
+
+## Validation Checklist
+
+For any schema block, verify:
+
+1. ✅ `@context` is `"https://schema.org"` (not http)
+2. ✅ `@type` is a valid, non-deprecated type
+3. ✅ All required properties are present
+4. ✅ Property values match expected data types
+5. ✅ No placeholder text (e.g., "[Business Name]")
+6. ✅ URLs are absolute, not relative
+7. ✅ Dates are in ISO 8601 format
+8. ✅ Images have valid URLs
+
+## Testing Tools
+
+- [Google Rich Results Test](https://search.google.com/test/rich-results)
+- [Schema.org Validator](https://validator.schema.org/)

package/skills/seo-aeo-landing-page-writer/SKILL.md

@@ -0,0 +1,97 @@
+---
+name: seo-aeo-landing-page-writer
+description: "Writes complete, structured landing pages optimized for SEO ranking, AEO citation, and visitor conversion / Penulis landing page terstruktur untuk peringkat SEO, sitasi AEO, dan konversi."
+author: "Roedy Rustam"
+---
+
+# SEO-AEO Landing Page Writer
+
+[English](#english) | [Bahasa Indonesia](#bahasa-indonesia)
+
+---
+
+<a name="english"></a>
+## English
+
+### Overview
+Generates a full, publish-ready landing page following a defined section order with SEO heading structure, AEO extraction blocks, FAQ section, comparison table, social proof, and conversion-focused CTAs. Every section serves a specific purpose in a narrative arc that moves the visitor from awareness to action.
+
+### Trigger Conditions
+Use this skill when:
+- Building a landing page for a new product or service.
+- An existing landing page needs a full SEO and AEO rewrite.
+- You need a page that can be cited by AI engines like Perplexity or ChatGPT.
+- You want conversion copy that leads with pain before pitching the product.
+
+### How It Works (Step-by-Step)
+1. **Map Inputs**: Extract product name, audience, primary keyword, pain points, features, benefits, USPs, social proof, and CTAs.
+2. **Write AEO Extraction Sentence**: Write one 25–40 word sentence that answers "What is [product]?" — standalone, no jargon, placed in a blockquote immediately after the H1. This is the sentence AI engines extract.
+3. **Follow the Narrative Arc**: Write sections in this exact order:
+   - Hero — H1 + AEO sentence + CTA
+   - Problem — Audience pain, no product mention yet
+   - Solution — Introduce product as the answer
+   - Features as Benefits — Table format
+   - Social Proof — Testimonials, logos, stats
+   - Mid-page CTA
+   - How It Works — Numbered steps
+   - Comparison — Table with honest competitor comparison
+   - FAQ — Minimum 6 entries, each under 50 words
+   - Trust Signals
+   - Final CTA
+
+### Best Practices & Pitfalls
+- **Empathy First**: Write the problem section before mentioning the product — empathy first.
+- **AEO Block**: Place the AEO extraction sentence in a blockquote immediately after H1.
+- **FAQ Limits**: Every FAQ answer must be under 50 words and self-contained.
+- **Tone**: Do not use "revolutionary", "game-changing", or "best-in-class".
+- **CTA**: Do not use "Submit" or "Click Here" as CTA button text.
+
+### Layout & UI Aesthetic (Monday.com Inspired)
+- Design the landing page with a modern, spacious aesthetic (see `monday-design-aesthetic` skill).
+- Use generous vertical padding (80-120px) between narrative sections.
+- Display features as benefits using large, rounded cards (12-16px radius) with subtle drop shadows.
+- Keep the background clean white or light gray, using a vibrant brand color for CTA buttons.
+
+
+---
+
+<a name="bahasa-indonesia"></a>
+## Bahasa Indonesia
+
+### Ringkasan
+Menghasilkan landing page lengkap yang siap dipublikasikan dengan struktur heading SEO, blok ekstraksi AEO, bagian FAQ, tabel perbandingan, bukti sosial (social proof), dan CTA yang berfokus pada konversi. Setiap bagian dirancang untuk memandu pengunjung dari pemahaman masalah menuju tindakan (pembelian/registrasi).
+
+### Kondisi Pemicu
+Gunakan skill ini ketika:
+- Membangun landing page untuk produk atau layanan baru.
+- Menulis ulang landing page yang ada untuk optimasi SEO dan AEO.
+- Memerlukan halaman yang dapat dengan mudah dikutip oleh mesin pencari AI (seperti Perplexity atau ChatGPT).
+- Membuat salinan konversi (conversion copy) yang menyoroti masalah pengguna sebelum memperkenalkan solusi.
+
+### Cara Kerja (Langkah demi Langkah)
+1. **Pemetaan Input**: Ekstrak nama produk, target audiens, kata kunci utama, keluhan utama pengguna, fitur, manfaat, USP, bukti sosial, dan CTA.
+2. **Tulis Kalimat Ekstraksi AEO**: Buat satu kalimat ringkas berisi 25–40 kata yang menjawab "Apa itu [produk]?" tanpa jargon, dan letakkan di dalam blockquote tepat setelah tag H1. Kalimat ini yang akan diekstraksi oleh mesin pencari AI.
+3. **Ikuti Alur Narasi**: Tulis bagian dalam urutan persis seperti berikut:
+   - Hero — H1 + Kalimat AEO + CTA
+   - Problem — Keluhan utama audiens (tanpa menyebutkan produk terlebih dahulu)
+   - Solution — Memperkenalkan produk sebagai jawaban masalah
+   - Features as Benefits — Tabel fitur dan manfaat
+   - Social Proof — Testimoni, logo klien, statistik
+   - Mid-page CTA — CTA tengah halaman
+   - How It Works — Langkah bernomor
+   - Comparison — Tabel perbandingan jujur dengan kompetitor
+   - FAQ — Minimal 6 entri, masing-masing di bawah 50 kata
+   - Trust Signals — Sinyal kepercayaan (garansi, enkripsi, dll.)
+   - Final CTA — CTA penutup
+
+### Praktik Terbaik & Hal yang Harus Dihindari
+- **Empati Utama**: Tulis bagian masalah terlebih dahulu sebelum menyebutkan produk untuk membangun empati.
+- **Batasan FAQ**: Setiap jawaban FAQ harus di bawah 50 kata, mandiri, dan mudah dipahami tanpa konteks eksternal.
+- **Bahasa Jujur**: Hindari kata-kata berlebihan seperti "revolutionary", "game-changing", or "best-in-class".
+- **CTA Spesifik**: Jangan gunakan teks tombol standar seperti "Submit" atau "Click Here". Gunakan teks berorientasi tindakan (misal: "Mulai Uji Coba Gratis").
+
+### Tata Letak & Estetika UI (Inspirasi Monday.com)
+- Desain landing page dengan estetika modern dan lapang (lihat skill `monday-design-aesthetic`).
+- Gunakan padding vertikal yang lega (80-120px) di antara setiap bagian narasi.
+- Tampilkan fitur sebagai manfaat (features as benefits) menggunakan kartu bersudut melengkung besar (radius 12-16px) dengan efek bayangan halus.
+- Jaga latar belakang tetap putih bersih atau abu-abu terang, menggunakan warna merek yang cerah untuk tombol CTA.

package/skills/seo-geo/SKILL.md

@@ -0,0 +1,188 @@
+---
+name: seo-geo
+description: "Generative Engine Optimization (GEO) for AI Overviews, ChatGPT, and Perplexity / Optimasi Mesin Pencari Generatif (GEO) untuk AI Overviews, ChatGPT, dan Perplexity."
+author: "Roedy Rustam"
+---
+
+# SEO-GEO Expert (2026 — AI Mode Edition)
+
+[English](#english) | [Bahasa Indonesia](#bahasa-indonesia)
+
+---
+
+<a name="english"></a>
+## English
+
+### Description
+Generative Engine Optimization (GEO) for the 2026 AI-powered search landscape — Google AI Mode (successor to AI Overviews), ChatGPT Search, Perplexity Pages, Claude.ai web search, and Grok search. Optimizes content to be cited, quoted, and recommended by AI search engines.
+
+### Trigger Conditions
+- Optimizing content to appear in Google AI Mode responses.
+- Getting cited in ChatGPT Search, Perplexity, Claude web, or Grok results.
+- Adding `/llms.txt` and structured data for AI indexing.
+- Measuring GEO performance (citation rate, AI snippet appearances).
+- Optimizing for "Answer Engine" queries (who, what, how, why questions).
+
+### The 2026 AI Search Landscape
+
+| AI Engine | Source | Citable Content | Key Signal |
+|---|---|---|---|
+| **Google AI Mode** | Google index + Gemini | Yes — with source links | E-E-A-T, structured data |
+| **ChatGPT Search** | Bing index + GPT-4o | Yes — with citations | Bing ranking, freshness |
+| **Perplexity** | Real-time web crawl | Yes — prominent citations | Authority, specificity |
+| **Perplexity Pages** | User-generated long-form | Yes — auto-cited | Depth, comprehensiveness |
+| **Claude Web** | Real-time search | Yes — with URLs | Clarity, structure |
+| **Grok** | X/Twitter + web | Yes | Real-time trending |
+
+### GEO Optimization Framework
+
+#### 1. E-E-A-T Signals (Experience, Expertise, Authoritativeness, Trust)
+Google AI Mode heavily weights E-E-A-T:
+- **Author credentials**: Add `<script type="application/ld+json">` `Person` schema with credentials.
+- **Publication dates**: Always include `datePublished` and `dateModified`.
+- **Primary sources**: Cite original research, government data, and peer-reviewed studies.
+- **About page**: Clear author bios with LinkedIn/professional links.
+
+#### 2. `/llms.txt` — AI Indexing Protocol
+Add an `/llms.txt` file at your domain root — the 2026 standard for telling AI crawlers what to index:
+```
+# /llms.txt
+# AI-friendly site map for LLM indexers
+
+## Product
+- [Features](/features): Complete feature list with pricing
+- [API Docs](/docs/api): REST API reference (JSON schemas included)
+- [Changelog](/changelog): Recent updates and releases
+
+## Content
+- [Blog](/blog): Technical articles on [topic]
+- [Tutorials](/tutorials): Step-by-step guides
+
+## Company
+- [About](/about): Team, mission, and founding story
+- [Pricing](/pricing): Current plan structure
+```
+
+#### 3. Structured Data for AI Comprehension
+```html
+<!-- FAQ Schema — AI engines love Q&A format -->
+<script type="application/ld+json">
+{
+  "@context": "https://schema.org",
+  "@type": "FAQPage",
+  "mainEntity": [{
+    "@type": "Question",
+    "name": "What is [Product]?",
+    "acceptedAnswer": {
+      "@type": "Answer",
+      "text": "Concise, factual answer in 2-3 sentences that AI can quote directly."
+    }
+  }]
+}
+</script>
+
+<!-- HowTo Schema — for instructional content -->
+<script type="application/ld+json">
+{
+  "@context": "https://schema.org",
+  "@type": "HowTo",
+  "name": "How to [Task]",
+  "step": [
+    { "@type": "HowToStep", "name": "Step 1", "text": "..." },
+    { "@type": "HowToStep", "name": "Step 2", "text": "..." }
+  ]
+}
+</script>
+```
+
+#### 4. Content Formatting for AI Citation
+AI engines prefer content that is:
+- **Scannable**: Short paragraphs, headers for each concept, bullet lists.
+- **Definitive**: Declarative statements ("X is Y") over hedging ("X might be Y").
+- **Self-contained**: Each section can stand alone as a quotable snippet.
+- **Cited**: Reference authoritative sources with links.
+- **Fresh**: Updated within the last 6 months — AI engines penalize stale content.
+
+#### 5. Perplexity Pages Optimization
+Perplexity Pages are AI-generated long-form content that cites sources. To be cited in Pages:
+- Cover a topic comprehensively (> 1500 words).
+- Use clear section headers (H2/H3 for each subtopic).
+- Include statistics, data, and specific numbers.
+- Have a clear "Summary" or "TL;DR" section at the top.
+
+#### 6. ChatGPT Search Optimization
+ChatGPT Search is powered by Bing — standard Bing SEO applies, plus:
+- Ensure Bing Webmaster Tools verified and sitemap submitted.
+- OpenGraph tags properly set (`og:title`, `og:description`, `og:image`).
+- Content freshness signals (update dates, news sitemaps).
+
+#### 7. GEO Performance Measurement
+```python
+# Automated GEO tracking
+import httpx
+
+async def check_ai_citation(brand: str, queries: list[str]):
+    for query in queries:
+        # Check Perplexity
+        perp_result = await query_perplexity(query)
+        cited = brand.lower() in perp_result.lower()
+        
+        # Log citation rate per query
+        await log_metric("geo_citation", {
+            "engine": "perplexity",
+            "query": query,
+            "cited": cited,
+        })
+```
+
+---
+
+<a name="bahasa-indonesia"></a>
+## Bahasa Indonesia
+
+### Deskripsi
+Generative Engine Optimization (GEO) untuk lanskap pencarian bertenaga AI tahun 2026 — Google AI Mode (penerus AI Overviews), ChatGPT Search, Perplexity Pages, pencarian web Claude.ai, dan pencarian Grok. Mengoptimalkan konten agar dikutip, disebut, dan direkomendasikan oleh mesin pencari AI.
+
+### Kondisi Pemicu
+- Mengoptimalkan konten agar muncul dalam respons Google AI Mode.
+- Mendapatkan kutipan di ChatGPT Search, Perplexity, Claude web, atau hasil Grok.
+- Menambahkan `/llms.txt` dan data terstruktur untuk pengindeksan AI.
+- Mengukur performa GEO (tingkat kutipan, kemunculan AI snippet).
+- Mengoptimalkan untuk kueri "Answer Engine" (pertanyaan siapa, apa, bagaimana, mengapa).
+
+### Lanskap Pencarian AI 2026
+
+| Mesin AI | Sumber | Konten yang Dapat Dikutip | Sinyal Kunci |
+|---|---|---|---|
+| **Google AI Mode** | Indeks Google + Gemini | Ya — dengan tautan sumber | E-E-A-T, data terstruktur |
+| **ChatGPT Search** | Indeks Bing + GPT-4o | Ya — dengan kutipan | Peringkat Bing, kesegaran |
+| **Perplexity** | Crawl web real-time | Ya — kutipan menonjol | Otoritas, spesifisitas |
+| **Perplexity Pages** | Konten panjang yang dibuat AI | Ya — dikutip otomatis | Kedalaman, komprehensivitas |
+| **Claude Web** | Pencarian real-time | Ya — dengan URL | Kejelasan, struktur |
+
+### Kerangka Optimasi GEO
+
+#### 1. Sinyal E-E-A-T
+Google AI Mode sangat mempertimbangkan E-E-A-T: kredensial penulis (schema `Person`), tanggal publikasi, kutipan dari sumber primer, dan halaman About yang jelas.
+
+#### 2. `/llms.txt` — Protokol Pengindeksan AI
+Tambahkan file `/llms.txt` di root domain Anda — standar 2026 untuk memberi tahu crawler AI apa yang harus diindeks. Mirip `robots.txt` tetapi untuk LLM.
+
+#### 3. Data Terstruktur untuk Pemahaman AI
+Gunakan schema `FAQPage` untuk konten Q&A dan `HowTo` untuk konten instruksional — format yang paling disukai mesin AI untuk dikutip.
+
+#### 4. Format Konten untuk Kutipan AI
+- Paragraf pendek, header untuk setiap konsep, daftar poin.
+- Pernyataan definitif ("X adalah Y") bukan hedging ("X mungkin Y").
+- Setiap seksi dapat berdiri sendiri sebagai snippet yang dapat dikutip.
+- Sertakan statistik, data, dan angka spesifik.
+- Diperbarui dalam 6 bulan terakhir.
+
+#### 5. Optimasi Perplexity Pages
+Tulis konten yang komprehensif (> 1500 kata), gunakan header seksi yang jelas, sertakan statistik dan angka spesifik, dan sertakan ringkasan "TL;DR" di bagian atas.
+
+#### 6. Optimasi ChatGPT Search
+ChatGPT Search didukung oleh Bing — SEO Bing standar berlaku. Verifikasi di Bing Webmaster Tools, submit sitemap, dan pastikan tag OpenGraph diatur dengan benar.
+
+#### 7. Pengukuran Performa GEO
+Lacak tingkat kutipan merek Anda per mesin AI per query secara otomatis untuk memantau visibilitas AI dari waktu ke waktu.

package/skills/session-handoff-resume/SKILL.md

@@ -0,0 +1,158 @@
+---
+name: session-handoff-resume
+description: "Skill to save ultra-compact project checkpoints and seamlessly resume work across accounts or new chat sessions with minimum token consumption / Skill untuk menyimpan checkpoint proyek yang super ringkas dan melanjutkan pekerjaan secara efisien saat ganti akun/sesi tanpa boros token."
+author: "Roedy Rustam"
+---
+
+# Session Handoff & Resume (2026 — Cross-Model Edition)
+
+[English](#english) | [Bahasa Indonesia](#bahasa-indonesia)
+
+---
+
+<a name="english"></a>
+## English
+
+### Description
+Saves ultra-compact project checkpoints and enables seamless work resumption across new chat sessions, different accounts, or different AI models (Gemini, Claude, GPT). Designed for minimum token consumption while preserving full context needed for continuation.
+
+### Trigger Conditions
+- Session is running very long and context window is nearly exhausted.
+- User wants to switch to a new chat session or different AI model.
+- User wants to save progress before closing and resume later.
+- User says "save checkpoint", "handoff", "save progress", "resume later".
+
+### Checkpoint Format (Ultra-Compact)
+
+When creating a checkpoint, output a compact YAML block under 200 tokens:
+
+```yaml
+# CHECKPOINT — [Project Name] — [Date]
+project: "[Project Name]"
+goal: "[One-line goal of the current session]"
+status: "[brief status: e.g., '60% — backend API done, frontend in progress']"
+stack: "[e.g., Next.js 15 + Hono + Supabase + Drizzle]"
+last_completed:
+  - "[Most recent task completed]"
+  - "[Second most recent]"
+next_tasks:
+  - "[Immediate next task]"
+  - "[Following task]"
+  - "[Following task]"
+blockers: "[Any known blockers or pending decisions, or 'none']"
+key_files:
+  - "[path/to/critical/file.ts]"
+  - "[path/to/another/file.ts]"
+env_needed: "[List of env vars needed, e.g., SUPABASE_URL, STRIPE_KEY]"
+notes: "[Any critical context that doesn't fit above]"
+```
+
+### Cross-Model Handoff Protocol (2026)
+
+Different AI models interpret context differently. Use this universal handoff prompt:
+
+```markdown
+## RESUME CONTEXT — [Project Name]
+
+You are resuming work on [Project Name]. Read this context carefully before taking any action.
+
+**Goal**: [One-sentence project goal]
+**Current Status**: [What's been done, what's pending]
+**Tech Stack**: [Stack details]
+**Repository**: [path or URL if available]
+
+**Completed**:
+- ✅ [Done item 1]
+- ✅ [Done item 2]
+
+**Next Tasks** (in priority order):
+1. [ ] [Immediate next task — be specific]
+2. [ ] [Second task]
+3. [ ] [Third task]
+
+**Critical Files to Read First**:
+- [file path] — [why it's critical]
+- [file path] — [why it's critical]
+
+**Known Constraints**:
+- [Constraint 1, e.g., "Use pnpm, not npm"]
+- [Constraint 2, e.g., "Admin panel must be on admin.domain.com"]
+
+**Do NOT**:
+- [Something the previous session explicitly avoided]
+- [Another anti-pattern to avoid]
+
+Start by reading the critical files listed above, then confirm you understand the current state before proceeding.
+```
+
+### Checkpoint Storage Locations
+
+| Destination | Use When |
+|---|---|
+| `CHECKPOINT.md` in project root | Primary — version-controlled |
+| `.agents/CHECKPOINT.md` | Alternative — agents-specific dir |
+| Copied to clipboard | Quick cross-session handoff |
+| Supabase notes table | Cloud-synced across devices |
+
+### Resume Protocol (For New Session)
+When the user provides a checkpoint to resume from:
+1. **Read all key files** listed in the checkpoint.
+2. **Verify current state** matches the checkpoint (check git status, recent commits).
+3. **Confirm understanding** — briefly state what you're resuming and what the first task is.
+4. **Proceed immediately** — don't ask unnecessary questions.
+
+### Automatic Checkpoint Triggers
+Proactively offer to create a checkpoint when:
+- User has been working for > 30 messages in a session.
+- The context window is > 70% consumed.
+- A major milestone is completed (e.g., "backend API is fully done").
+- The user says "I'll continue later".
+
+---
+
+<a name="bahasa-indonesia"></a>
+## Bahasa Indonesia
+
+### Deskripsi
+Menyimpan checkpoint proyek yang ultra-ringkas dan memungkinkan kelanjutan pekerjaan yang mulus di sesi chat baru, akun yang berbeda, atau model AI yang berbeda (Gemini, Claude, GPT). Dirancang untuk konsumsi token minimum sambil mempertahankan konteks penuh yang diperlukan untuk kelanjutan.
+
+### Kondisi Pemicu
+- Sesi berjalan sangat lama dan context window hampir habis.
+- Pengguna ingin beralih ke sesi chat baru atau model AI yang berbeda.
+- Pengguna ingin menyimpan progres sebelum menutup dan melanjutkan nanti.
+- Pengguna berkata "simpan checkpoint", "handoff", "simpan progres", "lanjut nanti".
+
+### Format Checkpoint (Ultra-Ringkas)
+
+Saat membuat checkpoint, keluarkan blok YAML ringkas di bawah 200 token dengan field: `project`, `goal`, `status`, `stack`, `last_completed`, `next_tasks`, `blockers`, `key_files`, `env_needed`, `notes`.
+
+### Protokol Handoff Lintas-Model (2026)
+
+Model AI yang berbeda menginterpretasikan konteks secara berbeda. Gunakan prompt handoff universal saat berpindah antar model (Gemini → Claude → GPT):
+- Sertakan goal satu kalimat, status saat ini, stack, daftar tugas yang selesai (✅), dan tugas berikutnya yang diprioritaskan.
+- Sertakan file kritis yang harus dibaca terlebih dahulu oleh model baru.
+- Sertakan kendala yang diketahui dan daftar "Jangan Lakukan".
+- Minta model baru untuk mengkonfirmasi pemahaman sebelum melanjutkan.
+
+### Lokasi Penyimpanan Checkpoint
+
+| Tujuan | Gunakan Saat |
+|---|---|
+| `CHECKPOINT.md` di root proyek | Utama — dikontrol versi |
+| `.agents/CHECKPOINT.md` | Alternatif — direktori agen |
+| Disalin ke clipboard | Handoff lintas sesi cepat |
+| Tabel catatan Supabase | Tersinkronisasi cloud antar perangkat |
+
+### Protokol Resume (Untuk Sesi Baru)
+Saat pengguna memberikan checkpoint untuk dilanjutkan:
+1. **Baca semua file kritis** yang tercantum dalam checkpoint.
+2. **Verifikasi status saat ini** cocok dengan checkpoint (periksa git status, commit terbaru).
+3. **Konfirmasi pemahaman** — nyatakan singkat apa yang dilanjutkan dan apa tugas pertama.
+4. **Lanjutkan segera** — jangan ajukan pertanyaan yang tidak perlu.
+
+### Pemicu Checkpoint Otomatis
+Tawarkan secara proaktif untuk membuat checkpoint saat:
+- Pengguna telah bekerja lebih dari 30 pesan dalam satu sesi.
+- Context window > 70% terpakai.
+- Milestone besar selesai.
+- Pengguna berkata "saya akan lanjutkan nanti".