Shared Workflow Generation

"Parallel workflow generation – one sub-agent per agent, no write conflicts."

When to Use This Skill

Use when:

• Running PRD Starter wizard Phase 11
• Generating workflow documentation for all agents

Use proactively:

• After all agents, skills, and files have been generated
• Before displaying final summary to user

---

Quick Start

<examples> Example 1: Launch all workflow generators ``` # Detect enabled agents from prd.json # Launch parallel sub-agents: Task("workflow-generator", { agent_name: "pm", output: "pm-coordinator.md" }) Task("workflow-generator", { agent_name: "developer", output: "developer.md" }) Task("workflow-generator", { agent_name: "qa", output: "qa.md" }) Task("devcycle-generator", { output: "development-cycle.md" }) ```

Example 2: Verify outputs

ls docs/workflows/
# Expected: pm-coordinator.md, developer.md, qa.md, development-cycle.md

Example 3: Generate index

# Agent Workflows Index

- **[Development Cycle](./development-cycle.md)**
- **[PM Coordinator](./pm-coordinator.md)**
- **[Developer](./developer.md)**
- **[QA Validator](./qa.md)**

</examples>

---

Parallel Execution Strategy

Each workflow document generated by separate sub-agent:

┌─────────────────────────────────────────────────────┐
│              WORKFLOW GENERATION ORCHESTRATOR       │
│  ┌─────────┐  ┌─────────┐  ┌─────────┐           │
│  │   pm    │  │ developer│  │   qa    │           │
│  └────┬────┘  └────┬────┘  └────┬────┘           │
│       │            │            │                  │
│       ▼            ▼            ▼                  │
│  ┌─────────┐  ┌─────────┐  ┌─────────┐          │
│  │pm-coord.md│  │developer.md│  │   qa.md  │          │
│  └─────────┘  └─────────┘  └─────────┘          │
│                                                     │
│  ┌─────────┐                                      │
│  │devcycle  │                                      │
│  └────┬────┘                                      │
│       ▼                                           │
│  ┌─────────┐                                      │
│ │dev-cycle.md│                                     │
│ └─────────┘                                      │
└─────────────────────────────────────────────────────┘

---

Your Process

1. Detect Enabled Agents

Read prd.json.agents for enabled agents.

2. Create Output Directory

mkdir -p docs/workflows

3. Launch Parallel Sub-Agents

Send ALL Task invocations in a single message:

Task("workflow-generator", {
  agent_name: "pm",
  output_file: "pm-coordinator.md",
  source_file: "agents/pm/AGENT.md"
})

Task("workflow-generator", {
  agent_name: "developer",
  output_file: "developer.md",
  source_file: "agents/developer/AGENT.md"
})

// ... one for each enabled agent

Task("devcycle-generator", {
  output_file: "development-cycle.md"
})

4. Wait for Completion

Each sub-agent confirms: Generated workflow documentation: {filename}

5. Verify Outputs

Expected files:

• pm-coordinator.md
• developer.md
• techartist.md
• qa.md
• gamedesigner.md
• development-cycle.md

6. Generate Index

Create docs/workflows/index.md with links to all workflows.

---

Agent Mapping

| Agent | Type | Output File | Source File | |-------|------|-------------|-------------| | PM | coordinator | pm-coordinator.md | agents/pm/AGENT.md | | Developer | worker | developer.md | agents/developer/AGENT.md | | Tech Artist | worker | techartist.md | agents/techartist/AGENT.md | | QA | worker | qa.md | agents/qa/AGENT.md | | Game Designer | worker | gamedesigner.md | agents/gamedesigner/AGENT.md | | - | orchestration | development-cycle.md | (from docs) |

---

Error Handling

If sub-agent fails:

1. Note the error
2. Continue with other agents
3. Report summary of succeeded/failed
4. Offer retry for failed workflows

---

Success Criteria

• All enabled agent workflow files created
• development-cycle.md created
• YAML frontmatter valid in all files
• Index file created with links