/work - Submit Project Request

SESSION ID: ${CLAUDE_SESSION_ID}

⚠️ IMPORTANT: Use this session ID for ALL role switches in this conversation:

bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" <role>.md

Welcome to X Company!

Architecture (v2.0)

Core Roles (READ, don't spawn): .claude/core/ (ceo.md, cto.md, hr.md, pm.md) Specialists (SPAWN these): .claude/agents/ (all specialists)

How It Works

When /work is invoked:

cd "$CLAUDE_PROJECT_DIR" && bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" ceo.md

Phase 1: CEO Approval
├── 1. 🎯 [CEO] → Read core/ceo.md → Analyze requirements
├── 2. 🎯 [CEO] → Ask HIGH-LEVEL questions if needed
└── 3. 🎯 [CEO] → Approve scope → Delegate to PM

Phase 2: Sprint 0 (Planning, then Foundation)
├── 4. 📋 [PM] → Read core/pm.md → Init project (bash init-project.sh)
├── 5. 📊 [BA] → Read core/ba.md → Ask DETAILED questions → Write SRS
├── 6. 📋 [PM] → ⭐ SPRINT 0 CHECKPOINTS ⭐ (user confirms decisions)
│       ├── Wireframes: Yes / No / External design?
│       ├── Tech Stack: OK / Modify?
│       └── Team: OK / Modify?
├── 7. 🏗️ [CTO] → Read core/cto.md → Finalize tech-stack.md + architecture.md
│       (File Blueprint + ⭐ Foundation Manifest) ── MUST precede step 9
├── 8. 👥 [HR] → Read core/hr.md → Finalize team.md + VERIFICATION
├── 9. 🎨 [UX] → (if agent designs) Spawn apple-ux-wireframer → ⛔ Phase 0.0 reads
│       tech-stack.md (tokens MUST match the chosen UI platform) → present trendy
│       directions → user picks → .project/design-system.md + primitives + wireframes
├── 10. 📋 [PM] → Phase Gate 1 Check → plan ALL sprints → user approves roadmap
└── 11. 📋 [PM] → ⭐ FOUNDATION BATCH (sprint-0.md — the only code in Sprint 0)
        ├── F1 structure per File Blueprint (app builds + starts)
        ├── F2 lint/format/type standards — FRONTEND *and* BACKEND, proven to reject
        ├── F3 design tokens compiled + BASE COMPONENT LAYER (all states)
        ├── F4 app/CONVENTIONS.md
        └── 🚦 validate-foundation.sh MUST exit 0 → only then Sprint 1

Phase 3: Development (Sprint 1+)
├── 11. 📋 [PM] → SPAWN specialists in PARALLEL
├── 12. 📋 [PM] → Track progress, resolve blockers
└── 13. 📋 [PM] → Report milestones to CEO

Phase 4: Bug Fix (Post-release)
├── 🎯 [CEO] → Detect bug report → Check active project → Confirm → Delegate to PM
└── 📋 [PM] → Read pm.md → See Bug Fix section → Read helpers/pm-bug-fix-flow.md
    ├── Step 1: Read project state
    ├── Step 2: Bug triage (PM interviews user — no role switch)
    ├── Step 3: Team check (PM assesses team — no role switch)
    ├── Step 4: Create bug fix sprint (create-sprint.sh)
    ├── Step 5: Present plan → wait for user approval
    └── Step 6: SPAWN specialists → Code review → QA

Key points:

• CEO approves, PM orchestrates
• PM "wears different hats" (CTO, HR, BA) when reading those roles
• Sprint 0 Checkpoints allow user to skip wireframes, modify tech stack, adjust team
• Sprint 0 Foundation is mandatory — structure, FE+BE code standards, and the base component layer are built and verified BEFORE any feature sprint. From Sprint 1 on, 2-5 agents build in parallel; anything not decided once here is invented N times
• Stack before design — the wireframer reads tech-stack.md first, so the design system is emitted in the target UI platform's own token dialect
• Role indicators (🎯🏗️👥📋📊🎨) are MANDATORY for each output
• HR verification checklist MUST be completed before spawning
• NO nested agent spawning - only one layer of actual agents (specialists)
• Monitor: Run /monitor start to view real-time dashboard (auto-tracking enabled)

⭐ Role Delegation Rules (MANDATORY)

CRITICAL: When switching roles, you MUST:

1. Bash echo to .claude/.active-core with session ID and target role filename
2. Read the target role file

Delegation Matrix

From	To	Action Required
CEO	PM	Bash(bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" pm.md) → Read(core/pm.md)
PM	CTO	Bash(bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" cto.md) → Read(core/cto.md)
PM	HR	Bash(bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" hr.md) → Read(core/hr.md)
PM	BA	Bash(bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" ba.md) → Read(core/ba.md)
BA	PM	Bash(bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" pm.md) → Read(core/pm.md)
CTO	PM	Bash(bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" pm.md) → Read(core/pm.md)
HR	PM	Bash(bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" pm.md) → Read(core/pm.md)

Correct Example

🎯 [CEO] → Project approved. Delegating to PM.
         ↓
[Bash: bash .claude/automation/set-active-core.sh "${CLAUDE_SESSION_ID}" pm.md]   ← REQUIRED!
[Read: .claude/core/pm.md]                                                         ← REQUIRED!
         ↓
📋 [PM] → Received handover from CEO...

Wrong Example

🎯 [CEO] Project approved. Delegating to PM.    ← ❌ Missing set-active-core.sh call!
📋 [PM] Starting project...                     ← ❌ Did NOT read pm.md!

Commands

Create New Project

/work "Build a social media app like Instagram"

Continue Project

/work continue
/work tiếp tục

Report Bugs

/work I found bugs: login fails on mobile, dashboard chart not rendering (+ screenshots)
/work trang settings bị lỗi, nút save không hoạt động

When user message relates to bugs/errors/issues but doesn't explicitly say "bug", CEO MUST ask to confirm: "It sounds like you're reporting bugs. Is that correct?"

Check Status

/work status