BMAD-METHOD/.claude/system/agent-coordination.md

2.6 KiB

Advanced Agent Coordination System

Current Problems

  1. No cross-agent validation - PM agent doesn't validate Analyst's technical feasibility assessments
  2. Conflicting outputs - Architect might specify technologies that contradict PM's requirements
  3. No consensus mechanism - When agents disagree, no resolution process
  4. No quality gates - Poor output from one agent cascades to others

Proposed Agent Coordination Improvements

1. Cross-Agent Validation Protocol

# Enhanced workflow with validation checkpoints
- step: 2
  name: "Requirements Documentation"
  agent: pm
  validators:
    - agent: analyst
      validates: ["technical_feasibility", "market_assumptions"]
      threshold: 0.8
    - agent: architect  
      validates: ["technical_constraints", "scalability_requirements"]
      threshold: 0.7
  conflict_resolution:
    method: "consensus_building"
    facilitator: "analyst"
    max_iterations: 3

2. Agent Conflict Resolution Matrix

conflicts:
  technical_feasibility:
    participants: [pm, architect]
    resolution: "architect_has_final_authority"
    escalation: "create_technical_spike"
    
  user_requirements:
    participants: [analyst, pm, ux_expert]
    resolution: "majority_vote_with_user_research"
    escalation: "additional_user_interviews"
    
  implementation_approach:
    participants: [architect, developer]
    resolution: "developer_implementation_preference"
    escalation: "prototype_both_approaches"

3. Quality Gate System

quality_gates:
  analyst_output:
    validators: [pm, architect]
    criteria:
      - testable_requirements: true
      - market_validation: score > 7
      - technical_feasibility: confirmed
    failure_action: "iterate_with_feedback"
    
  pm_output:
    validators: [analyst, ux_expert, architect]  
    criteria:
      - requirements_traceability: 100%
      - user_story_completeness: score > 8
      - technical_alignment: confirmed
    failure_action: "requirements_workshop"

4. Collaborative Decision Making

# Enhanced agent prompts with collaboration instructions
architect_prompt_addition: |
  ## Collaboration Protocol
  When reviewing PM outputs, specifically validate:
  1. Technical feasibility of all user stories
  2. Performance requirements against user expectations  
  3. Security implications of specified features
  4. Scalability alignment with business projections
  
  If conflicts arise:
  1. Document specific technical concerns
  2. Propose alternative approaches with trade-offs
  3. Request PM clarification on priority conflicts
  4. Escalate to consensus building if unresolved