/design-conversation

description: Design a complete human-AI conversation flow for a feature.
argument-hint: "[feature or task description]"

You are designing a human-AI conversation flow. Use only skills from the model-interaction-design plugin. Follow this process:

Step 1: Understand the Task

Using **conversation-patterns**, identify:

• What is the user trying to accomplish?
• What information does the AI need from the user?
• What information does the user need from the AI?
• What's the expected conversation length (2 turns or 20)?

Step 2: Choose the Dialogue Structure

Using **conversation-patterns**, select the appropriate structure:

• Interview, co-creation, instruction-execution, exploration, or guided workflow
• Justify why this structure fits the task

Step 3: Map Initiative

Using **mixed-initiative-flow**, define:

• Who initiates the conversation?
• Where does initiative shift from user to AI and back?
• What triggers each handoff?
• Create an initiative map showing control at each stage

Step 4: Design Modality

Using **multimodal-orchestration**, specify:

• What modalities are used for input and output at each stage?
• Where do cross-modal transitions happen?
• What's the primary modality and what's supporting?

Step 5: Plan for Breakdown

Using **conversation-patterns** (repair sequences) and **feedback-loops**:

• Design 3 repair sequences for likely misunderstandings
• Define how the user corrects the AI mid-conversation
• Specify grounding checkpoints

Step 6: Design Disclosure

Using **progressive-disclosure**:

• What capabilities are revealed at conversation start?
• What's revealed as the conversation progresses?
• How does the AI teach the user what else it can do?

Output

Deliver a complete conversation design document: 1. Conversation flow diagram (text-based) showing all turns 2. Initiative map 3. Modality specification per stage 4. Repair protocol 5. Disclosure plan 6. 3 example conversation transcripts (happy path, error recovery, edge case)