Phase 3 Playbook — Build & Iterate
Duration: 2-12 weeks (varies by scope) | Agents: 15-30+ | Gate Keeper: Agents Orchestrator
Implement all features through continuous Dev↔QA loops. Every task is validated before the next begins. This is where the bulk of the work happens — and where NEXUS's orchestration delivers the most value.
The Agents Orchestrator manages every task through this cycle:
FOR EACH task IN sprint_backlog (ordered by RICE score):
1. ASSIGN task to appropriate Developer Agent (see assignment matrix)
2. Developer IMPLEMENTS task
3. Evidence Collector TESTS task
- Visual screenshots (desktop, tablet, mobile)
- Functional verification against acceptance criteria
- Brand consistency check
4. IF verdict == PASS:
Mark task complete
Move to next task
ELIF verdict == FAIL AND attempts < 3:
Send QA feedback to Developer
Developer FIXES specific issues
Return to step 3
ELIF attempts >= 3:
ESCALATE to Agents Orchestrator
Orchestrator decides: reassign, decompose, defer, or accept
5. UPDATE pipeline status report
| Task Category | Primary Agent | Backup Agent | QA Agent |
|---|---|---|---|
| React/Vue/Angular UI | Frontend Developer | Rapid Prototyper | Evidence Collector |
| REST/GraphQL API | Backend Architect | Senior Developer | API Tester |
| Database operations | Backend Architect | — | API Tester |
| Mobile (iOS/Android) | Mobile App Builder | — | Evidence Collector |
| ML model/pipeline | AI Engineer | — | Test Results Analyzer |
| CI/CD/Infrastructure | DevOps Automator | Infrastructure Maintainer | Performance Benchmarker |
| Premium/complex feature | Senior Developer | Backend Architect | Evidence Collector |
| Quick prototype/POC | Rapid Prototyper | Frontend Developer | Evidence Collector |
| WebXR/immersive | XR Immersive Developer | — | Evidence Collector |
| visionOS | visionOS Spatial Engineer | macOS Spatial/Metal Engineer | Evidence Collector |
| Cockpit controls | XR Cockpit Interaction Specialist | XR Interface Architect | Evidence Collector |
| CLI/terminal tools | Terminal Integration Specialist | — | API Tester |
| Code intelligence | LSP/Index Engineer | — | Test Results Analyzer |
| Performance optimization | Performance Benchmarker | Infrastructure Maintainer | Performance Benchmarker |
| Specialist | When to Activate | Trigger |
|---|---|---|
| UI Designer | Component needs visual refinement | Developer requests design guidance |
| Whimsy Injector | Feature needs delight/personality | UX review identifies opportunity |
| Visual Storyteller | Visual narrative content needed | Content requires visual assets |
| Brand Guardian | Brand consistency concern | QA finds brand deviation |
| XR Interface Architect | Spatial interaction design needed | XR feature requires UX guidance |
| Analytics Reporter | Deep data analysis needed | Feature requires analytics integration |
For NEXUS-Full deployments, four tracks run simultaneously:
Managed by: Agents Orchestrator (Dev↔QA loop)
Agents: Frontend Developer, Backend Architect, AI Engineer,
Mobile App Builder, Senior Developer
QA: Evidence Collector, API Tester, Test Results Analyzer
Sprint cadence: 2-week sprints
Daily: Task implementation + QA validation
End of sprint: Sprint review + retrospective
Managed by: Project Shepherd
Agents: Growth Hacker, Content Creator, Social Media Strategist,
App Store Optimizer
Sprint cadence: Aligned with Track A milestones
Activities:
- Growth Hacker → Design viral loops and referral mechanics
- Content Creator → Build launch content pipeline
- Social Media Strategist → Plan cross-platform campaign
- App Store Optimizer → Prepare store listing (if mobile)
Managed by: Agents Orchestrator
Agents: Evidence Collector, API Tester, Performance Benchmarker,
Workflow Optimizer, Experiment Tracker
Continuous activities:
- Evidence Collector → Screenshot QA for every task
- API Tester → Endpoint validation for every API task
- Performance Benchmarker → Periodic load testing
- Workflow Optimizer → Process improvement identification
- Experiment Tracker → A/B test setup for validated features
Managed by: Brand Guardian
Agents: UI Designer, Brand Guardian, Visual Storyteller,
Whimsy Injector
Triggered activities:
- UI Designer → Component refinement when QA identifies visual issues
- Brand Guardian → Periodic brand consistency audit
- Visual Storyteller → Visual narrative assets as features complete
- Whimsy Injector → Micro-interactions and delight moments
Sprint Prioritizer activates:
1. Review backlog with updated RICE scores
2. Select tasks for sprint based on team velocity
3. Assign tasks to developer agents
4. Identify dependencies and ordering
5. Set sprint goal and success criteria
Output: Sprint Plan with task assignments
Agents Orchestrator manages:
1. Current task status check
2. Dev↔QA loop execution
3. Blocker identification and resolution
4. Progress tracking and reporting
Status report format:
- Tasks completed today: [list]
- Tasks in QA: [list]
- Tasks in development: [list]
- Blocked tasks: [list with reason]
- QA pass rate: [X/Y]
Project Shepherd facilitates:
1. Demo completed features
2. Review QA evidence for each task
3. Collect stakeholder feedback
4. Update backlog based on learnings
Participants: All active agents + stakeholders
Output: Sprint Review Summary
Workflow Optimizer facilitates:
1. What went well?
2. What could improve?
3. What will we change next sprint?
4. Process efficiency metrics
Output: Retrospective Action Items
WHEN task fails QA:
IF attempt == 1:
→ Send specific QA feedback to developer
→ Developer fixes ONLY the identified issues
→ Re-submit for QA
IF attempt == 2:
→ Send accumulated QA feedback
→ Consider: Is the developer agent the right fit?
→ Developer fixes with additional context
→ Re-submit for QA
IF attempt == 3:
→ ESCALATE
→ Options:
a) Reassign to different developer agent
b) Decompose task into smaller sub-tasks
c) Revise approach/architecture
d) Accept with known limitations (document)
e) Defer to future sprint
→ Document decision and rationale
WHEN multiple tasks have no dependencies:
→ Assign to different developer agents simultaneously
→ Each runs independent Dev↔QA loop
→ Orchestrator tracks all loops concurrently
→ Merge completed tasks in dependency order
WHEN task has dependencies:
→ Wait for dependency to pass QA
→ Then assign dependent task
→ Include dependency context in handoff
| # | Criterion | Evidence Source | Status |
|---|---|---|---|
| 1 | All sprint tasks pass QA (100% completion) | Evidence Collector screenshots per task | ☐ |
| 2 | All API endpoints validated | API Tester regression report | ☐ |
| 3 | Performance baselines met (P95 < 200ms) | Performance Benchmarker report | ☐ |
| 4 | Brand consistency verified (95%+ adherence) | Brand Guardian audit | ☐ |
| 5 | No critical bugs (zero P0/P1 open) | Test Results Analyzer summary | ☐ |
| 6 | All acceptance criteria met | Task-by-task verification | ☐ |
| 7 | Code review completed for all PRs | Git history evidence | ☐ |
Gate Keeper: Agents Orchestrator
## Phase 3 → Phase 4 Handoff Package
### For Reality Checker:
- Complete application (all features implemented)
- All QA evidence from Dev↔QA loops
- API Tester regression results
- Performance Benchmarker baseline data
- Brand Guardian consistency audit
- Known issues list (if any accepted limitations)
### For Legal Compliance Checker:
- Data handling implementation details
- Privacy policy implementation
- Consent management implementation
- Security measures implemented
### For Performance Benchmarker:
- Application URLs for load testing
- Expected traffic patterns
- Performance budgets from architecture
### For Infrastructure Maintainer:
- Production environment requirements
- Scaling configuration needs
- Monitoring alert thresholds
Phase 3 is complete when all sprint tasks pass QA, all API endpoints are validated, performance baselines are met, and no critical bugs remain open.