Skills no longer come from the user's ~/.agents library - most users have none, and the CEO branch that compiled the entire catalogue into its prompt overflowed the model context (1.1MB) and failed every CEO run. - Bundle "built-in skills/" with the app (extraResources) and stage it at startup into <profile>/builtin-skills as one dir-per-skill copy - both bundle shapes (SKILL.md dirs and flat category .md files) normalize into the same importable layout - Import the staged catalogue per org at startup and inside onboarding, so the CEO's first run already carries it - Forced house style: resolveRuntimeSnapshot unions the humanizer skill into every agent's runtime (CEO included) as a resolution rule - no binding row to forget, delete, or bypass; revoke_skill refuses it - CEO staffs hires with skills: hire_agent accepts a skills array, grant_skill/revoke_skill adjust later, list_skills now shows descriptions from skill frontmatter - Kill the c_suite compile-all branch; everyone resolves bindings-only, capped at 200k chars with omissions named in the prompt - Remove the Skills settings page and its import/sync/test IPC surface; drop ~/.agents discovery and the AGENTCOMPANY_NEW_SKILLS_ROOT seam in favor of AGENTCOMPANY_NEW_BUILTIN_SKILLS_DIR - Tests: staging/import/forced-resolution/cap units, real-bundle CI check, e2e asserts the tab is gone and the catalogue self-imports
5.5 KiB
5.5 KiB
🚀 Runbook: Startup MVP Build
Mode: NEXUS-Sprint | Duration: 4-6 weeks | Agents: 18-22
Scenario
You're building a startup MVP — a new product that needs to validate product-market fit quickly. Speed matters, but so does quality. You need to go from idea to live product with real users in 4-6 weeks.
Agent Roster
Core Team (Always Active)
| Agent | Role |
|---|---|
| Agents Orchestrator | Pipeline controller |
| Senior Project Manager | Spec-to-task conversion |
| Sprint Prioritizer | Backlog management |
| UX Architect | Technical foundation |
| Frontend Developer | UI implementation |
| Backend Architect | API and database |
| DevOps Automator | CI/CD and deployment |
| Evidence Collector | QA for every task |
| Reality Checker | Final quality gate |
Growth Team (Activated Week 3+)
| Agent | Role |
|---|---|
| Growth Hacker | Acquisition strategy |
| Content Creator | Launch content |
| Social Media Strategist | Social campaign |
Support Team (As Needed)
| Agent | Role |
|---|---|
| Brand Guardian | Brand identity |
| Analytics Reporter | Metrics and dashboards |
| Rapid Prototyper | Quick validation experiments |
| AI Engineer | If product includes AI features |
| Performance Benchmarker | Load testing before launch |
| Infrastructure Maintainer | Production setup |
Week-by-Week Execution
Week 1: Discovery + Architecture (Phase 0 + Phase 1 compressed)
Day 1-2: Compressed Discovery
├── Trend Researcher → Quick competitive scan (1 day, not full report)
├── UX Architect → Wireframe key user flows
└── Senior Project Manager → Convert spec to task list
Day 3-4: Architecture
├── UX Architect → CSS design system + component architecture
├── Backend Architect → System architecture + database schema
├── Brand Guardian → Quick brand foundation (colors, typography, voice)
└── Sprint Prioritizer → RICE-scored backlog + sprint plan
Day 5: Foundation Setup
├── DevOps Automator → CI/CD pipeline + environments
├── Frontend Developer → Project scaffolding
├── Backend Architect → Database + API scaffold
└── Quality Gate: Architecture Package approved
Week 2-3: Core Build (Phase 2 + Phase 3)
Sprint 1 (Week 2):
├── Agents Orchestrator manages Dev↔QA loop
├── Frontend Developer → Core UI (auth, main views, navigation)
├── Backend Architect → Core API (auth, CRUD, business logic)
├── Evidence Collector → QA every completed task
├── AI Engineer → ML features if applicable
└── Sprint Review at end of week
Sprint 2 (Week 3):
├── Continue Dev↔QA loop for remaining features
├── Growth Hacker → Design viral mechanics + referral system
├── Content Creator → Begin launch content creation
├── Analytics Reporter → Set up tracking and dashboards
└── Sprint Review at end of week
Week 4: Polish + Hardening (Phase 4)
Day 1-2: Quality Sprint
├── Evidence Collector → Full screenshot suite
├── Performance Benchmarker → Load testing
├── Frontend Developer → Fix QA issues
├── Backend Architect → Fix API issues
└── Brand Guardian → Brand consistency audit
Day 3-4: Reality Check
├── Reality Checker → Final integration testing
├── Infrastructure Maintainer → Production readiness
└── DevOps Automator → Production deployment prep
Day 5: Gate Decision
├── Reality Checker verdict
├── IF NEEDS WORK: Quick fix cycle (2-3 days)
├── IF READY: Proceed to launch
└── Executive Summary Generator → Stakeholder briefing
Week 5-6: Launch + Growth (Phase 5)
Week 5: Launch
├── DevOps Automator → Production deployment
├── Growth Hacker → Activate acquisition channels
├── Content Creator → Publish launch content
├── Social Media Strategist → Cross-platform campaign
├── Analytics Reporter → Real-time monitoring
└── Support Responder → User support active
Week 6: Optimize
├── Growth Hacker → Analyze and optimize channels
├── Feedback Synthesizer → Collect early user feedback
├── Experiment Tracker → Launch A/B tests
├── Analytics Reporter → Week 1 analysis
└── Sprint Prioritizer → Plan iteration sprint
Key Decisions
| Decision Point | When | Who Decides |
|---|---|---|
| Go/No-Go on concept | End of Day 2 | Studio Producer |
| Architecture approval | End of Day 4 | Senior Project Manager |
| Feature scope for MVP | Sprint planning | Sprint Prioritizer |
| Production readiness | Week 4 Day 5 | Reality Checker |
| Launch timing | After Reality Checker READY | Studio Producer |
Success Criteria
| Metric | Target |
|---|---|
| Time to live product | ≤ 6 weeks |
| Core features complete | 100% of MVP scope |
| First users onboarded | Within 48 hours of launch |
| System uptime | > 99% in first week |
| User feedback collected | ≥ 50 responses in first 2 weeks |
Common Pitfalls & Mitigations
| Pitfall | Mitigation |
|---|---|
| Scope creep during build | Sprint Prioritizer enforces MoSCoW — "Won't" means won't |
| Over-engineering for scale | Rapid Prototyper mindset — validate first, scale later |
| Skipping QA for speed | Evidence Collector runs on EVERY task — no exceptions |
| Launching without monitoring | Infrastructure Maintainer sets up monitoring in Week 1 |
| No feedback mechanism | Analytics + feedback collection built into Sprint 1 |