feat: Domain/Architecture Feedback Gates + Phase 0 Agent Creation

[lkronecker13]

Summary

This PR significantly enhances the Project Initialization Workflow with two major features:

1. Domain & Architecture Feedback Gates - Ensure implementations are aligned with business outcomes and not over-engineered
2. Phase 0 Agent Creation - Allow users to create project-specific specialists when none are found

Problem

• Implementers could produce technically correct but over-engineered or misaligned solutions
• No mechanism to validate business alignment during implementation phases
• Users with no pre-existing specialist agents had no easy path to create them

Solution

New Roles

Role	Purpose
domain-expert	Business alignment, PRD compliance, user value validation
architecture-reviewer	Over-engineering detection, ADR compliance, pattern validation

Phase 3.4 & 4.3: Domain/Architecture Feedback

IMPLEMENT → CODE REVIEW → DOMAIN REVIEW → ARCHITECTURE REVIEW → APPROVAL GATE
                              ↓                    ↓
                         FIX ISSUES           SIMPLIFY

• Sequential reviews: Domain first, then architecture (can consider domain findings)
• Blocking: Both must pass before approval gate
• Max 2 cycles: Escalate to user if issues persist

Over-Engineering Signals Detected

• Abstractions without current use cases
• Premature optimization
• Features beyond Must-Have scope
• Configuration/flexibility not requested
• Layers/indirection without clear benefit

Phase 0 Enhancements

• 0.1 Scan & Match: Existing discovery + domain expert inference from PRODUCT.md
• 0.2 Gap Analysis: User chooses Create / Generalist / Hybrid
• 0.3 Specialist Creation:
  • Analyze seeds for domain context
  • Fetch best practices via claude-code-guide
  • Draft agents, present for approval
  • Skills configuration
  • Restart handling with state persistence

Domain Expert Selection

Inferred from context/PRODUCT.md:

Domain Keywords	Suggested Agent
ML, model, training	data-science-modeler
API, backend, database	backend-engineer
Agent, autonomous, LLM	ai-engineer
(No match)	Ask user / product-manager fallback

Files Changed

• workflows/PROJECT_INIT_WORKFLOW.md - Major enhancements (+645 lines)
• Removed unused: .serena/.gitignore, CHANGELOG.md, research/EDA.ipynb

Test Plan

• Run workflow with missing specialists → verify agent creation flow
• Implement a feature → verify domain/architecture feedback triggers
• Intentionally over-engineer → verify architecture review catches it
• Test restart recovery via Workflow State section

🤖 Generated with Claude Code