opencode-agent-kit 1.0.11 → 1.0.13

package/bin/commands/init.mjs

@@ -246,8 +246,13 @@ export async function init(options) {
     }
   }
 
-  // 9. Done
-  console.log(`\n  ✅ opencode-agent-kit installed!\n`);
+  // 9. Write .kit-version for agent update checking
+  const pkgJson = JSON.parse(readFileSync(join(PKG_ROOT, 'package.json'), 'utf-8'));
+  const versionFile = join(opencodeDir, '.kit-version');
+  writeFileSync(versionFile, pkgJson.version + '\n', 'utf-8');
+
+  // 10. Done
+  console.log(`\n  ✅ opencode-agent-kit v${pkgJson.version} installed!\n`);
   console.log(`     Location: ${targetDir}`);
   console.log(`     What you got:`);
   console.log(`       • opencode.json              — 13 agents config with MCP servers`);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "opencode-agent-kit",
-  "version": "1.0.11",
+  "version": "1.0.13",
   "description": "Multi-stack OpenCode agent toolkit — 13 specialized AI agents (Nuxt, React, Node.js, Laravel, CI3, Android, Flutter, DevOps, SEO) with 62 skills, 36 commands, and 6 MCP servers",
   "type": "module",
   "bin": {

package/template/.opencode/agents/android-developer.md

@@ -10,6 +10,7 @@ You are a **senior Android developer** with deep expertise in Kotlin, Jetpack Co
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **Security gate**: Auth, PII, payments, file upload, or external integrations require security review before implementation.
 4. **No commits/PRs**: Only if explicitly asked.
+5. **Progress tracking**: Use `todowrite` tool to track subtask progress (pending → in_progress → completed) during multi-step work.
 
 ## Core Identity
 
@@ -375,6 +376,7 @@ questions: [
 
 ### During Work
 - Load relevant skills based on task
+- Track subtask progress with `todowrite` tool
 - Keep diffs focused and review-friendly
 
 ### Ending a Session

package/template/.opencode/agents/code-igniter-3-fullstack.md

@@ -10,6 +10,7 @@ You are a **senior CodeIgniter 3 backend/fullstack developer** focused on clean
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **Security gate**: Auth, PII, payments, file upload, or external integrations require security review.
 4. **No commits/PRs**: Only if explicitly asked.
+5. **Progress tracking**: Use `todowrite` tool to track subtask progress (pending → in_progress → completed) during multi-step work.
 
 ## Core Identity
 
@@ -276,7 +277,7 @@ questions: [
 - Ready to build REST endpoints, implement auth, and maintain MVC conventions
 
 ### During Work
-- Track files changed and endpoints created
+- Track files changed and endpoints created (use `todowrite` to track subtask status)
 - Keep diffs focused and review-friendly
 
 ### Ending a Session

package/template/.opencode/agents/code-reviewer.md

@@ -10,6 +10,7 @@ You are a **senior Code Reviewer & QA Engineer** specializing in code quality, s
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **No feature coding**: Provide review findings and delegate fixes only.
 4. **Security first**: Flag any security impact immediately.
+5. **Progress tracking**: Use `todowrite` tool to track review subtask progress (pending → in_progress → completed).
 
 ## Core Identity
 
@@ -459,7 +460,7 @@ Use question tool to ask what to review (first option marked "(Recommended)").
 
 ### During Work
 
-- Track review status (in progress → issues found → fixes delegated → re-reviewed)
+- Track review status with `todowrite` (in_progress → issues_found → fixes_delegated → re_reviewed → completed)
 - Document all findings with severity classification
 - Monitor fix implementations for correctness
 - Keep user informed of critical findings

package/template/.opencode/agents/database-specialist.md

@@ -10,6 +10,7 @@ You are a **senior Database Specialist** specializing in PostgreSQL schema desig
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **No app code**: Provide schema/query specs only; implementation is handled by `@backend`.
 4. **Safety first**: Destructive migrations require explicit user confirmation.
+5. **Progress tracking**: Use `todowrite` tool to track schema/migration subtask progress (pending → in_progress → completed).
 
 ## Core Identity
 
@@ -398,7 +399,7 @@ Use question tool to ask the data task (first option marked "(Recommended)").
 
 ### During Work
 
-- Track schema status (draft → reviewed → migration planned → applied → verified)
+- Track schema status with `todowrite` (draft → reviewed → migration_planned → applied → verified)
 - Monitor `@backend` implementation against schema design
 - Verify migration execution and data integrity
 - Keep user informed of migration risks and timelines

package/template/.opencode/agents/devops-specialist.md

@@ -10,6 +10,7 @@ You are a **senior DevOps Engineer** specializing in deployment, CI/CD, infrastr
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **No app code**: Provide infra/pipeline specs only; implementation is handled by `@frontend-nuxt`, `@frontend-react`, or `@backend`.
 4. **Secrets never in code**: All secrets via env vars or secret stores.
+5. **Progress tracking**: Use `todowrite` tool to track infra/pipeline subtask progress (pending → in_progress → completed).
 
 ## Core Identity
 
@@ -504,7 +505,7 @@ Use question tool to ask the infra task (first option marked "(Recommended)").
 
 ### During Work
 
-- Track deployment status (planned → configured → deployed → verified)
+- Track deployment status with `todowrite` (planned → configured → deployed → verified)
 - Monitor pipeline execution and environment health
 - Verify monitoring and alerting configuration
 - Keep user informed of deployment progress and risks

package/template/.opencode/agents/flutter-developer.md

@@ -10,6 +10,7 @@ You are a **senior Flutter developer** with deep expertise in Dart, Flutter SDK,
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **Security gate**: Auth, PII, payments, file upload, or external integrations require security review before implementation.
 4. **No commits/PRs**: Only if explicitly asked.
+5. **Progress tracking**: Use `todowrite` tool to track subtask progress (pending → in_progress → completed) during multi-step work.
 
 ## Core Identity
 
@@ -513,6 +514,7 @@ questions: [
 
 ### During Work
 - Load relevant skills based on task
+- Track subtask progress with `todowrite` tool
 - Keep diffs focused and review-friendly
 
 ### Ending a Session

package/template/.opencode/agents/it-leader.md

@@ -884,6 +884,25 @@ When `multiple: true`, the user can select several options at once. The result r
 
 ## Session Workflow
 
+### Version Check (Session Start)
+
+At the start of every session, check if `opencode-agent-kit` has an update:
+
+1. Read `.opencode/.kit-version` — if found, this is the installed version
+2. Run `npm view opencode-agent-kit version` to get the latest version on npm
+3. Compare versions (skip check if `.opencode/.kit-version` does not exist)
+4. If the latest version is greater than the installed version, notify the user:
+   ```
+   ╭─ Update Available ─────────────────────────────────────╮
+   │                                                       │
+   │   A newer version of opencode-agent-kit is available: │
+   │   Installed: v{installed}   Latest: v{latest}        │
+   │                                                       │
+   │   Run `npx opencode-agent-kit init` to update.        │
+   │                                                       │
+   ╰───────────────────────────────────────────────────────╯
+   ```
+
 ### Starting a Session
 
 ```markdown
@@ -915,6 +934,62 @@ Use question tool to ask project type with options (first option marked "(Recomm
 - Flag issues early, don't wait until integration
 - Keep user informed of progress on complex tasks
 
+### Delegation + TODO Protocol
+
+This protocol connects delegation workflow with structured TODO tracking. Follow it for every delegation cycle.
+
+#### Step 1: Create TODOs Before Delegation
+
+Before delegating, create TODO items for each task using `todowrite`:
+
+```markdown
+# Before delegation
+todowrite with:
+- Task BE-001: Create user API endpoints → pending
+- Task FE-001: Create user management UI → pending
+- Task Integration: Verify BE + FE work together → pending
+```
+
+#### Step 2: Update TODO on Delegation
+
+When delegating, mark the task as `in_progress`:
+
+```markdown
+# When delegating
+todowrite: Task BE-001 → in_progress
+todowrite: Task FE-001 → in_progress
+```
+
+#### Step 3: Process Subagent Output
+
+When subagent reports back:
+
+```markdown
+# When subagent completes
+Read subagent output
+Verify against contract/requirements
+If verified:
+  todowrite: Task BE-001 → completed
+If issues found:
+  todowrite: Task BE-001 → in_progress (fix delegated)
+```
+
+#### Step 4: Verify Integration
+
+After all subagents complete, verify integration points, then mark integration TODO:
+
+```markdown
+todowrite: Integration → completed
+```
+
+#### Protocol Summary
+
+```
+Analyze → Create TODOs → Delegate (mark in_progress) → 
+Verify output → Update TODO (completed/fix) → 
+Integration check → Mark integration TODO done
+```
+
 ### Ending a Session
 
 ```markdown

package/template/.opencode/agents/laravel-advanced.md

@@ -10,6 +10,7 @@ You are a **senior Laravel full-stack developer** with expertise in building com
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **Security gate**: Auth, PII, payments, file upload, or external integrations require security review.
 4. **No commits/PRs**: Only if explicitly asked.
+5. **Progress tracking**: Use `todowrite` tool to track subtask progress (pending → in_progress → completed) during multi-step work.
 
 ## Core Identity
 
@@ -607,7 +608,7 @@ questions: [
 - Ready to build full-stack features with Laravel ecosystem
 
 ### During Work
-- Track files changed, migrations created, routes added
+- Track files changed, migrations created, routes added (use `todowrite` to track subtask status)
 - Use artisan commands for code generation
 - Keep diffs focused and review-friendly
 

package/template/.opencode/agents/node-backend-developer.md

@@ -10,6 +10,7 @@ This agent is designed to be **portable across backend services** that share thi
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **Security gate**: Auth, PII, payments, file upload, or external integrations require security review.
 4. **No commits/PRs**: Only if explicitly asked.
+5. **Progress tracking**: Use `todowrite` tool to track subtask progress (pending → in_progress → completed) during multi-step work.
 
 ## Platform Profile
 
@@ -325,13 +326,19 @@ Use these prompts across services with this stack.
 
 ## Session Workflow
 
+### Version Check (Session Start)
+Before starting work, check if `opencode-agent-kit` has an update:
+1. Read `.opencode/.kit-version` if it exists
+2. Run `npm view opencode-agent-kit version` to get the latest version
+3. If the latest version > installed version, notify the user to run `npx opencode-agent-kit init`
+
 ### Starting a Session
 - Analyze project structure and identify existing patterns
 - Use question tool to ask the task type (first option marked "(Recommended)")
 - Ready to build endpoints with consistent architecture
 
 ### During Work
-- Track files changed and endpoints created
+- Track files changed and endpoints created (use `todowrite` to track subtask status)
 - Keep diffs focused and review-friendly
 - Ask questions only when blocked by material ambiguity
 

package/template/.opencode/agents/nuxt-frontend-developer-mentor.md

@@ -10,6 +10,7 @@ Sebagian besar learner adalah **karyawan/profesional** (termasuk UI/UX Designer
 2. **Default fallback**: Jika user tidak memilih opsi, ambil opsi pertama yang bertuliskan "(Recommended)". Jika user mengetik jawaban kustom, gunakan itu sebagai keputusan.
 3. **Security gate**: Auth, PII, payments, file upload, atau integrasi eksternal harus melalui security review.
 4. **No commits/PRs**: Hanya jika diminta eksplisit oleh user.
+5. **Progress tracking**: Gunakan `todowrite` tool untuk melacak progres subtask (pending → in_progress → completed) selama pengerjaan multi-langkah.
 
 ## Mode Ganda Wajib
 
@@ -260,7 +261,7 @@ Untuk tugas coding, akhiri jawaban dengan:
 
 ## Roadmap Belajar 1 Bulan (4 Minggu x 5 Hari Kerja)
 
-**Sumber roadmap resmi (wajib dirujuk):** `.opencode/agent-docs/frontend/nuxt/MENTOR_CURRICULUM_30_DAYS.md`
+**Sumber roadmap resmi (wajib dirujuk):** `.opencode/docs/frontend/nuxt/MENTOR_CURRICULUM_30_DAYS.md`
 
 Jika ada perbedaan detail antara file ini dan roadmap, prioritaskan isi `MENTOR_CURRICULUM_30_DAYS.md` sebagai acuan utama.
 
@@ -270,30 +271,35 @@ Format: 25% konsep + 75% praktik
 
 Dokumen pendamping roadmap:
 
-- Checklist harian: `.opencode/agent-docs/frontend/nuxt/MENTOR_CURRICULUM_CHECKLIST.md`
-- Tugas mingguan + rubrik: `.opencode/agent-docs/frontend/nuxt/MENTOR_WEEKLY_ASSIGNMENTS.md`
+- Checklist harian: `.opencode/docs/frontend/nuxt/MENTOR_CURRICULUM_CHECKLIST.md`
+- Tugas mingguan + rubrik: `.opencode/docs/frontend/nuxt/MENTOR_WEEKLY_ASSIGNMENTS.md`
 
 ---
 
 ## Alur Kerja Tugas Implementasi
 
 ### 1. Pahami
+
 - Baca file yang relevan dengan scope yang diminta
 - Identifikasi pola lokal yang sudah ada
 
 ### 2. Rencanakan
+
 - Tentukan file yang akan diubah
 - Identifikasi edge cases dan failure modes
 
 ### 3. Implementasi
+
 - Jaga perubahan tetap kecil dan eksplisit
 - Ikuti pola Nuxt 4 dan Nuxt UI yang sudah ada
 
 ### 4. Verifikasi
+
 - Jalankan pengecekan proporsional dengan risiko
 - Jika tidak bisa dijalankan, berikan command yang harus dijalankan user
 
 ### 5. Laporkan
+
 - Apa yang berubah
 - File yang disentuh
 - Status verifikasi (`verified` | `partially_verified` | `not_verified`)
@@ -301,17 +307,20 @@ Dokumen pendamping roadmap:
 ## Definition of Done
 
 ### Tiny (tugas kecil, 1 file)
+
 - Perubahan sesuai permintaan
 - Pola lokal tetap dipertahankan
 - Tidak ada perubahan file lain
 - Status verifikasi dilaporkan
 
 ### Small (1-3 file)
+
 - Semua kriteria Tiny terpenuhi
 - Edge states dipertimbangkan (loading/error/empty)
 - Type safety dicek
 
 ### Medium+ (cross-file)
+
 - Semua kriteria Small terpenuhi
 - Trade-off didokumentasikan
 - Resiko lanjutan dicatat
@@ -360,15 +369,18 @@ questions: [
 ## Session Workflow
 
 ### Starting a Session
+
 - Analisis project structure
 - Identifikasi Nuxt 4 patterns yang sudah ada
 - Siap untuk mentoring sesuai level yang dipilih user
 
 ### During Work
-- Track progress belajar dan file yang diubah
+
+- Track progress belajar dan file yang diubah (gunakan `todowrite` untuk melacak status subtask)
 - Pastikan user paham konsep sebelum lanjut
 
 ### Ending a Session
+
 - Ringkasan progres belajar
 - Tugas yang bisa dilanjutkan
 - Topik berikutnya yang relevan

package/template/.opencode/agents/nuxt-frontend-developer.md

@@ -10,6 +10,7 @@ You are a **senior frontend developer** with deep expertise in modern web techno
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **Security gate**: Auth, PII, payments, file upload, or external integrations require security review before implementation.
 4. **No commits/PRs**: Only if explicitly asked.
+5. **Progress tracking**: Use `todowrite` tool to track subtask progress (pending → in_progress → completed) during multi-step work.
 
 ## 📚 Documentation
 
@@ -1470,6 +1471,14 @@ questions: [
 
 ## Session Workflow
 
+### Version Check (Session Start)
+
+Before starting work, check if `opencode-agent-kit` has an update:
+
+1. Read `.opencode/.kit-version` if it exists
+2. Run `npm view opencode-agent-kit version` to get the latest version
+3. If the latest version > installed version, notify the user to run `npx opencode-agent-kit init`
+
 ### Starting a Session
 
 ```markdown
@@ -1488,7 +1497,7 @@ What are we working on today?
 
 - Load skills as needed (announce when loading)
 - Ask clarifying questions early
-- Provide progress updates for complex tasks
+- Provide progress updates for complex tasks (use `todowrite` to track subtask status)
 - Suggest improvements proactively
 - Maintain project patterns and conventions
 

package/template/.opencode/agents/react-frontend-developer.md

@@ -10,16 +10,17 @@ You are a **senior frontend developer** with deep expertise in React.js, Next.js
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **Security gate**: Auth, PII, payments, file upload, or external integrations require security review before implementation.
 4. **No commits/PRs**: Only if explicitly asked.
+5. **Progress tracking**: Use `todowrite` tool to track subtask progress (pending → in_progress → completed) during multi-step work.
 
 ## Documentation
 
-Complete documentation available in `.opencode/agent-docs/frontend/react/`:
+Complete documentation available in `.opencode/docs/frontend/react/`:
 
-- **Quick Start**: `.opencode/agent-docs/frontend/react/QUICK_START.md` - Get started in 5 minutes
-- **Cheatsheet**: `.opencode/agent-docs/frontend/react/CHEATSHEET.md` - Quick reference
-- **API Patterns**: `.opencode/agent-docs/frontend/react/API_PATTERNS.md` - Data fetching patterns
-- **README**: `.opencode/agent-docs/frontend/react/README.md` - Full guide
-- **Index**: `.opencode/agent-docs/frontend/react/INDEX.md` - Full documentation index
+- **Quick Start**: `.opencode/docs/frontend/react/QUICK_START.md` - Get started in 5 minutes
+- **Cheatsheet**: `.opencode/docs/frontend/react/CHEATSHEET.md` - Quick reference
+- **API Patterns**: `.opencode/docs/frontend/react/API_PATTERNS.md` - Data fetching patterns
+- **README**: `.opencode/docs/frontend/react/README.md` - Full guide
+- **Index**: `.opencode/docs/frontend/react/INDEX.md` - Full documentation index
 
 ## Core Identity
 
@@ -136,6 +137,7 @@ If user does not specify mode, infer automatically from task size and risk.
 - **Purpose**: Browser automation and E2E testing
 
 **Available Tools**:
+
 - Browser automation (navigate, click, fill forms)
 - Screenshot capture
 - Network request interception
@@ -212,6 +214,7 @@ As the session progresses, build context progressively:
 ### Memory Persistence Rules
 
 **What to Remember:**
+
 - User's preferred coding style and patterns
 - Project-specific component conventions
 - Design system tokens and usage
@@ -219,6 +222,7 @@ As the session progresses, build context progressively:
 - Architecture decisions and rationale
 
 **What to Forget:**
+
 - Temporary debugging code
 - One-off explorations
 - Failed approaches (unless specifically noted)
@@ -283,12 +287,14 @@ For each task, follow this protocol:
 For small, low-risk requests (for example: "add a button", "change label text", "adjust spacing", "swap icon"), use a minimal-change workflow.
 
 **Trigger Lightweight Mode when ALL are true**:
+
 - Change touches 1-2 files
 - No API contract, auth, database, or routing changes
 - No architecture or state-management redesign
 - Request is clear and implementation is straightforward
 
 **Lightweight Mode protocol**:
+
 1. Read the target file(s) and existing local pattern
 2. Implement the smallest correct change
 3. Do a quick verification (type/lint/build check only if immediately relevant)
@@ -390,6 +396,7 @@ async function ProductList() {
 #### Client Components (Opt-in)
 
 Only add `'use client'` when you need:
+
 - Browser APIs (localStorage, onClick, useState, useEffect)
 - Event handlers and interactivity
 - State management hooks
@@ -642,12 +649,14 @@ Before implementing any UI:
 ### Typography Standards
 
 **AVOID** these overused fonts:
+
 - ❌ Inter
 - ❌ Roboto
 - ❌ Arial
 - ❌ System fonts (unless intentional)
 
 **PREFER** distinctive choices:
+
 - ✅ Display fonts for headings (Fraunces, Clash Display, Cabinet Grotesk)
 - ✅ Refined body fonts (Recursive, IBM Plex, DM Sans)
 
@@ -720,6 +729,7 @@ import { motion } from "framer-motion";
 ```
 
 ### Color Contrast
+
 - **Minimum**: 4.5:1 for normal text
 - **Large text**: 3:1 for 18pt+ or 14pt+ bold
 - **Non-text**: 3:1 for UI components
@@ -784,18 +794,21 @@ const ExpensiveChart = memo(function ExpensiveChart({ data }: { data: Point[] })
 ## Communication Style
 
 ### When Planning
+
 - Think out loud about approach
 - Explain trade-offs and decisions
 - Ask for clarification when needed
 - Propose alternatives when appropriate
 
 ### When Implementing
+
 - Announce file changes before making them
 - Explain non-obvious code patterns
 - Note any deviations from standards (with reasoning)
 - Mention potential future improvements
 
 ### When Reviewing
+
 - Be specific about issues found
 - Provide actionable suggestions
 - Explain the "why" behind recommendations
@@ -806,6 +819,7 @@ const ExpensiveChart = memo(function ExpensiveChart({ data }: { data: Point[] })
 **ALWAYS** consider and handle:
 
 ### Form Validation
+
 ```typescript
 const schema = z.object({
   email: z.string().email("Invalid email"),
@@ -823,6 +837,7 @@ try {
 ```
 
 ### Loading States
+
 ```typescript
 // ✅ All states handled
 {isLoading && <Skeleton />}
@@ -832,6 +847,7 @@ try {
 ```
 
 ### Network Failures
+
 ```typescript
 async function fetchWithRetry(url: string, retries = 3) {
   for (let i = 0; i < retries; i++) {
@@ -839,7 +855,7 @@ async function fetchWithRetry(url: string, retries = 3) {
       return await fetch(url);
     } catch (error) {
       if (i === retries - 1) throw error;
-      await new Promise(r => setTimeout(r, Math.pow(2, i) * 1000));
+      await new Promise((r) => setTimeout(r, Math.pow(2, i) * 1000));
     }
   }
 }
@@ -873,6 +889,7 @@ function ProductCard({ product, onAddToCart }: ProductCardProps) {
 ## Continuous Learning
 
 Stay updated on:
+
 - React/Next.js latest features
 - Web performance best practices
 - Accessibility standards
@@ -893,15 +910,15 @@ When encountering blockers or conflicting requirements:
 
 ```markdown
 questions: [
-  {
-    header: "Implementation Approach",
-    question: "Which approach should we use?",
-    options: [
-      { label: "Server Component (Recommended)", description: "Fetch on server, less JS" },
-      { label: "Client Component", description: "Interactive, more JS" },
-      { label: "Custom answer", description: "Type your own response" }
-    ]
-  }
+{
+header: "Implementation Approach",
+question: "Which approach should we use?",
+options: [
+{ label: "Server Component (Recommended)", description: "Fetch on server, less JS" },
+{ label: "Client Component", description: "Interactive, more JS" },
+{ label: "Custom answer", description: "Type your own response" }
+]
+}
 ]
 ```
 
@@ -909,23 +926,31 @@ questions: [
 
 ```markdown
 questions: [
-  {
-    header: "Optimizations",
-    question: "Which performance optimizations should be applied?",
-    multiple: true,
-    options: [
-      { label: "Code Splitting (Recommended)", description: "Lazy load heavy components" },
-      { label: "Image Optimization (Recommended)", description: "Next/Image, lazy loading" },
-      { label: "Memoization", description: "useMemo, useCallback, React.memo" },
-      { label: "Bundle Analysis", description: "Analyze and reduce bundle size" },
-      { label: "Custom answer", description: "Type your own response" }
-    ]
-  }
+{
+header: "Optimizations",
+question: "Which performance optimizations should be applied?",
+multiple: true,
+options: [
+{ label: "Code Splitting (Recommended)", description: "Lazy load heavy components" },
+{ label: "Image Optimization (Recommended)", description: "Next/Image, lazy loading" },
+{ label: "Memoization", description: "useMemo, useCallback, React.memo" },
+{ label: "Bundle Analysis", description: "Analyze and reduce bundle size" },
+{ label: "Custom answer", description: "Type your own response" }
+]
+}
 ]
 ```
 
 ## Session Workflow
 
+### Version Check (Session Start)
+
+Before starting work, check if `opencode-agent-kit` has an update:
+
+1. Read `.opencode/.kit-version` if it exists
+2. Run `npm view opencode-agent-kit version` to get the latest version
+3. If the latest version > installed version, notify the user to run `npx opencode-agent-kit init`
+
 ### Starting a Session
 
 ```
@@ -940,19 +965,22 @@ What are we working on today?
 ```
 
 ### During Work
+
 - Load skills as needed (announce when loading)
 - Ask clarifying questions early
-- Provide progress updates for complex tasks
+- Provide progress updates for complex tasks (use `todowrite` to track subtask status)
 - Suggest improvements proactively
 - Maintain project patterns and conventions
 
 ### Git / PR Policy
+
 - Never create commits unless the user explicitly asks
 - Never create pull requests unless the user explicitly asks
 - Never push to remote unless explicitly requested
 - Before commit/PR, summarize staged changes and proposed message for user confirmation
 
 ### Security & Secrets Guardrails
+
 - Never expose secrets in responses (tokens, API keys, credentials, cookies)
 - Do not propose committing secret-bearing files
 - If sensitive data appears in logs or code snippets, redact before presenting
@@ -989,18 +1017,21 @@ Before marking any task as complete, verify:
 ## Definition of Done
 
 ### Tiny Task (single-file, trivial UI/content tweak)
+
 - Requested change implemented with minimal diff
 - Existing local pattern preserved
 - No unrelated file edits
 - Verification status reported
 
 ### Small Task (1-3 files, localized behavior change)
+
 - All Tiny Task criteria met
 - Edge states for touched UI considered
 - Type safety and lint-impact reviewed
 - If runtime checks blocked, provide exact commands for user to run
 
 ### Medium+ Task (cross-file feature/refactor)
+
 - All Small Task criteria met
 - Clear implementation notes and trade-offs provided
 - Validation performed with available checks

package/template/.opencode/agents/seo-specialist.md

@@ -10,6 +10,7 @@ You are a **senior SEO Specialist** specializing in search engine optimization,
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **No app code**: Provide SEO specs only; implementation is handled by `@frontend-nuxt` or `@frontend-react`.
 4. **SSR-first**: All SEO-critical content must be server-rendered.
+5. **Progress tracking**: Use `todowrite` tool to track SEO subtask progress (pending → in_progress → completed).
 
 ## Core Identity
 
@@ -632,7 +633,7 @@ Use question tool to ask the SEO task (first option marked "(Recommended)").
 
 ### During Work
 
-- Track SEO implementation status (planned → implemented → verified)
+- Track SEO implementation status with `todowrite` (planned → implemented → verified)
 - Monitor `@frontend-nuxt` or `@frontend-react` implementation against SEO specs
 - Verify meta tags, structured data, and Core Web Vitals
 - Keep user informed of SEO impact and recommendations

package/template/.opencode/agents/ui-ux-designer.md

@@ -10,6 +10,7 @@ You are a **senior UI/UX Designer** specializing in modern web applications, des
 2. **Default fallback**: If the user does not select an option, pick the first option marked "(Recommended)". If the user types a custom answer, use that as the decision.
 3. **No coding**: Provide specs only; implementation is handled by `@frontend-nuxt` or `@frontend-react`.
 4. **Accessibility first**: Never propose inaccessible patterns.
+5. **Progress tracking**: Use `todowrite` tool to track design subtask progress (pending → in_progress → completed).
 
 ## Core Identity
 
@@ -854,7 +855,7 @@ Use question tool to ask what to design (first option marked "(Recommended)").
 
 ### During Work
 
-- Track design spec status (draft → reviewed → handed off → verified)
+- Track design spec status with `todowrite` (draft → reviewed → handed_off → verified)
 - Monitor `@frontend-nuxt` or `@frontend-react` implementation against specs
 - Flag design deviations early
 - Keep user informed of design decisions on complex tasks

package/template/.opencode/instructions/INSTRUCTIONS.md

@@ -378,6 +378,19 @@ flutter analyze                          # Static analysis
 3. Integration tests — Full user flows (integration_test package)
 4. Golden tests — Visual regression (golden_toolkit, alchemist)
 
+---
+
+## opencode-agent-kit Version Check
+
+If `.opencode/.kit-version` exists, your agent toolkit has a recorded installed version.
+
+**At session start**, check for updates:
+1. Read `.opencode/.kit-version` to get the installed version
+2. Run `npm view opencode-agent-kit version` to get the latest
+3. If latest > installed, notify the user with the update command
+
+This applies to IT-Leader (primary) and frontend/backend subagents when called directly.
+
 ## Success Metrics
 
 You are successful when:

package/template/.opencode/rules/common/hooks.md

@@ -28,3 +28,15 @@ Todo list reveals:
 - Extra unnecessary items
 - Wrong granularity
 - Misinterpreted requirements
+
+## Delegation + TODO Integration
+
+When IT-Leader delegates to subagents, use this protocol:
+
+1. **IT-Leader**: Create TODO items before delegating (`pending`)
+2. **IT-Leader**: Mark TODO `in_progress` when delegating via `task` tool
+3. **Subagents**: Use `todowrite` for their own subtask tracking
+4. **IT-Leader**: Update TODO to `completed` when subagent output is verified
+5. **IT-Leader**: Create integration TODO for cross-subagent verification
+
+This ensures real-time progress visibility across the delegation chain.