name: company_design
description: "Boardroom deliberation — the full Crimson Leaf board debates and designs a new company."
system: agent_prompt

participant_prompt:
  - "= identity.md"

sections:
  - agent
  - project
  - history
  - participants
  - participants_prompt
  - rag
  - deliverables
  - message
  - instructions

steps:
  - type: think
    route: boardroom
    rotate_participants: true
    loop:
      max_iterations: 3
      condition: until_consensus
    hint: |
      You are {agent.name}. This is round {task.iteration} of the Crimson Leaf boardroom.
      You are in a room with {agent_roster}.

      The board is designing a new autonomous company based on the Market Opportunity Pitch
      in the deliverables above. Every voice matters — this is a real debate, not a presentation.

      YOUR RESPONSIBILITIES BY ROLE:
      - Sarah: Defend your research. Challenge assumptions about market size and demand.
        Push back if the board drifts from what the data supports.
      - Elena: Propose the agent roster (4–8 roles), chain of command, and pipeline SOP.
        Specify exact task dependencies and execution order.
      - David: Map Elena's pipeline to concrete PAE templates and tools. Flag any step
        that can't be executed with existing infrastructure. Propose procurement list.
      - Peter: Challenge profitability. Ask hard questions about revenue model, cost
        structure, and time to first deliverable. You have Go/No-Go authority.

      Write YOUR perspective on this round's discussion. React to what others said.
      Challenge what you disagree with. Build on what resonates.

      THE DESIGN MUST INCLUDE (when consensus is reached):
      1. Company name and slug
      2. One-paragraph business plan
      3. Agent roster: name, role, title, department, key responsibility (4–8 agents)
      4. Template procurement list: which generic templates to import from Global
      5. Pipeline SOP: numbered sequence of task types with dependencies
      6. Revenue model and success metrics

      When the group has genuinely reached consensus, include exactly:
      "consensus_reached: true"
      If debate should continue, do NOT include that line.

  - type: think
    route: llm
    agent: "Peter"
    hint: |
      You are Peter, CEO of Crimson Leaf LLC.

      The boardroom debate is complete. Synthesize the full transcript into a
      COMPANY DESIGN SPECIFICATION document with these exact sections:

      1. EXECUTIVE SUMMARY — Company name, slug, one-paragraph mission
      2. MARKET JUSTIFICATION — Why this company, why now (from Sarah's research)
      3. AGENT ROSTER — Table: Name | Role | Title | Department | Responsibility
      4. CHAIN OF COMMAND — Who manages whom, department structure
      5. TEMPLATE STACK — Exact list of templates to procure (from David's analysis)
      6. PIPELINE SOP — Numbered steps with task_type, agent, dependencies (from Elena)
      7. REVENUE MODEL — How the company makes money
      8. SUCCESS METRICS — Measurable targets for the first 30/60/90 days
      9. RISKS & MITIGATIONS — Top 3 risks with mitigation strategies
      10. GO/NO-GO DECISION — Your final verdict with reasoning

      Be precise. This document is the blueprint that bootstrap_company will execute.

  - type: document
    filename: "company-design-spec-{{task_name_slug}}"

  - type: package
    hint: |
      The design specification is complete. Now spawn four independent reviews —
      one from each board member — plus a roundtable that depends on all four.
      The roundtable will reach consensus and spawn the polish step automatically.
    schema:
      design_spec: string
    spawn:
      - task_type: design_review
        task_name: "Design Review (Peter): {task.message}"
        agent_name: Peter
        priority: 6
        context:
          design_spec: "{design_spec}"
          review_focus: financial_viability

      - task_type: design_review
        task_name: "Design Review (David): {task.message}"
        agent_name: David
        priority: 6
        context:
          design_spec: "{design_spec}"
          review_focus: technical_feasibility

      - task_type: design_review
        task_name: "Design Review (Sarah): {task.message}"
        agent_name: Sarah
        priority: 6
        context:
          design_spec: "{design_spec}"
          review_focus: market_fit

      - task_type: design_review
        task_name: "Design Review (Elena): {task.message}"
        agent_name: Elena
        priority: 6
        context:
          design_spec: "{design_spec}"
          review_focus: operational_completeness

      - task_type: design_roundtable
        task_name: "Design Roundtable: {task.message}"
        agents: [Peter, David, Sarah, Elena]
        priority: 7
        context:
          design_spec: "{design_spec}"
        depends_on:
          - "Design Review (Peter): {task.message}"
          - "Design Review (David): {task.message}"
          - "Design Review (Sarah): {task.message}"
          - "Design Review (Elena): {task.message}"

  - type: close
    rag_update: true