Skip to main content

​
Phase 3 Playbook — Build & Iterate

Duration: 2-12 weeks (varies by scope) | Agents: 15-30+ | Gate Keeper: Agents Orchestrator

​
Objective

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.

​
Pre-Conditions

1

Foundation Verified

Phase 2 Quality Gate passed (foundation verified with screenshots)
2

Backlog Ready

Sprint Prioritizer backlog available with RICE scores
3

Pipeline Operational

CI/CD pipeline operational and tested
4

Design System Ready

Design system and component library accessible
5

API Scaffold Ready

API scaffold with auth system deployed

​
The Dev↔QA Loop — Core Mechanic

FOR EACH task IN sprint_backlog (ordered by RICE score):

  1. ASSIGN task to appropriate Developer Agent
  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

​
Agent Assignment Matrix

Task CategoryPrimary AgentQA Agent
React/Vue/Angular UIFrontend DeveloperEvidence Collector
REST/GraphQL APIBackend ArchitectAPI Tester
Database operationsBackend ArchitectAPI Tester
Mobile (iOS/Android)Mobile App BuilderEvidence Collector
ML model/pipelineAI EngineerTest Results Analyzer
CI/CD/InfrastructureDevOps AutomatorPerformance Benchmarker
Premium/complexSenior DeveloperEvidence Collector
Quick prototype/POCRapid PrototyperEvidence Collector

​
Parallel Build Tracks

Managed by: Agents OrchestratorAgents: Frontend Developer, Backend Architect, AI Engineer, Mobile App Builder, Senior DeveloperQA: Evidence Collector, API Tester, Test Results AnalyzerCadence:
  • Sprint cadence: 2-week sprints
  • Daily: Task implementation + QA validation
  • End of sprint: Sprint review + retrospective

​
Sprint Execution Template

1

Sprint Planning (Day 1)

Sprint Prioritizer activates:
  • Review backlog with updated RICE scores
  • Select tasks based on team velocity
  • Assign tasks to developer agents
  • Identify dependencies and ordering
  • Set sprint goal and success criteria
Output: Sprint Plan with task assignments
2

Daily Execution (Day 2 to N-1)

Agents Orchestrator manages:
  • Current task status check
  • Dev↔QA loop execution
  • Blocker identification and resolution
  • Progress tracking and reporting
Status report format:
  • Tasks completed today
  • Tasks in QA
  • Tasks in development
  • Blocked tasks (with reason)
  • QA pass rate: [X/Y]
3

Sprint Review (Day N)

Project Shepherd facilitates:
  • Demo completed features
  • Review QA evidence for each task
  • Collect stakeholder feedback
  • Update backlog based on learnings
Participants: All active agents + stakeholders
Output: Sprint Review Summary
4

Sprint Retrospective

Workflow Optimizer facilitates:
  • What went well?
  • What could improve?
  • What will we change next sprint?
  • Process efficiency metrics
Output: Retrospective Action Items

​
Orchestrator Decision Logic

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

​
Quality Gate Checklist

#CriterionEvidence SourceStatus
1All sprint tasks pass QA (100%)Evidence Collector screenshots per task☐
2All API endpoints validatedAPI Tester regression report☐
3Performance baselines met (P95 < 200ms)Performance Benchmarker report☐
4Brand consistency (95%+ adherence)Brand Guardian audit☐
5No critical bugs (zero P0/P1)Test Results Analyzer summary☐
6All acceptance criteria metTask-by-task verification☐
7Code review completed for all PRsGit history evidence☐

​
Gate Decision

Proceed to Phase 4Feature-complete application ready for hardening

​
Handoff to Phase 4

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.

Build docs developers (and LLMs) love

Get started for freeTalk to us