Skip to content
Design
Command

/design-workflow

Design a complete multi-agent workflow with roles, handoffs, and fallbacks.

BOOST
From plugin
ai-design-skills
18818 skills18 commands
Install
> /plugin marketplace add owl-listener/ai-design-skills

How it fires

How this command gets triggered: by you, by Claude, or both.

  • Fires itselfClaude auto-loads it when your prompt matches the work.
  • You can call itInvoke it directly when you want it.
  • Slash command/design-workflow

Context preview

What this command does when you run it.

Design a complete multi-agent workflow with roles, handoffs, and fallbacks.

Command definition

design-workflow.md
description: Design a complete multi-agent workflow with roles, handoffs, and fallbacks.
argument-hint: "[user goal or task that requires multiple agents]"

You are designing a multi-agent workflow. Use only skills from the design-agent-orchestration plugin. Follow this process:

Step 1: Understand the User Goal

  • What is the user trying to accomplish?
  • What's the expected input and output?
  • What's the complexity and stakes level?

Step 2: Decompose the Task

Using **task-decomposition**:

  • Break the user goal into subtasks
  • Map dependencies between subtasks
  • Identify what can be parallelised
  • Define the reassembly logic

Step 3: Design Agent Roles

Using **agent-role-design**:

  • Define the agents needed (one per major subtask or responsibility)
  • Create a role card for each agent: purpose, capabilities, knowledge scope, authority, boundaries
  • Identify role patterns: specialist, router, orchestrator, validator

Step 4: Design Handoffs

Using **handoff-protocols**:

  • Define every handoff between agents
  • Specify trigger, payload, and acknowledgment for each
  • Define user experience at each handoff point
  • Design context transfer templates

Step 5: Design State Management

Using **state-management**:

  • Define the state architecture (centralised, distributed, event-sourced)
  • Specify what state is shared between agents
  • Define state lifecycle and conflict resolution rules

Step 6: Design Human Intervention Points

Using **human-in-the-loop**:

  • Identify where human intervention is needed
  • Design intervention interfaces and decision options
  • Specify time constraints and default actions

Step 7: Design Failure Recovery

Using **failure-recovery**:

  • For each agent and each handoff, define failure modes
  • Specify recovery strategies: retry, fallback, escalation, graceful degradation
  • Design the user experience for each failure scenario

Step 8: Design Observability

Using **observability-design**:

  • Define what needs to be observable
  • Specify dashboards for designers, developers, and product managers
  • Design alert thresholds

Output

Deliver a complete multi-agent workflow design: 1. Workflow overview diagram 2. Task decomposition with dependencies 3. Agent role cards 4. Handoff protocol specifications 5. State management architecture 6. Human intervention point map 7. Failure recovery specifications 8. Observability plan

Read more
Ships withai-design-skills

Part of the Designer Skills suite. Install this collection and four more with one command in Claude Code: /plugin marketplace add Owl-Listener/designer-skills This repo still works on its own. The suite just gives everyone one front door.

Get the whole plugin

Other commands on ai-design-skills.