Facilitate

---
description: "Interactive refinement and collaborative exploration with Facilitator agent"
allowed-tools: ['Read', 'Edit', 'Task']
argument-hint: "[topic|task-id] [--mode explore|refine|validate]"
---

# Command: /cf:facilitate

## Usage

/cf:facilitate [topic|task-id] [--mode explore|refine|validate]

## Parameters

- `[topic|task-id]`: **Optional** - Topic to explore or task ID to refine (default: current focus from activeContext.md)
- `[--mode explore]`: **Optional** - Open-ended exploration mode for ambiguous topics
- `[--mode refine]`: **Optional** - Refinement mode for improving existing plans or specs (default)
- `[--mode validate]`: **Optional** - Validation mode for confirming understanding and catching gaps

---

## Purpose

Engage the Facilitator agent for human-in-the-loop interaction:
1. Interactive exploration of ambiguous or complex topics
2. Refinement of plans, requirements, or specifications through dialogue
3. Gap identification and resolution before implementation
4. Validation of understanding and assumptions
5. Collaborative decision-making for high-impact choices

---

## Prerequisites

**Memory bank must be initialized**. Run `/cf:init` first if needed.

**Topic or task context**. Works best when there's something to explore or refine.

---

## When to Use

**Exploration mode** (--mode explore):
- Ambiguous requirements that need clarification
- New feature ideas that aren't fully formed
- Architecture decisions with multiple viable options
- Open-ended "what if" scenarios
- Early-stage project discovery

**Refinement mode** (--mode refine, default):
- Existing plans that need improvement
- Task specifications that feel incomplete
- Requirements that have gaps or ambiguities
- Technical approaches with unresolved questions
- Plans from /cf:plan that need iteration

**Validation mode** (--mode validate):
- Confirming shared understanding before implementation
- Catching assumptions that need validation
- Edge case discovery
- Risk identification
- Pre-implementation sanity check

---

## Process

### Step 1: Verify Memory Bank Exists

Check if `memory-bank/` directory exists:

**If NOT EXISTS**:

⚠️ Memory Bank Not Initialized

Run: /cf:init

**Stop execution.**

---

### Step 2: Determine Facilitation Context

**No topic/task-id provided**:
- Read `activeContext.md` to determine current focus
- Facilitate around current work
- Default to refine mode

**Topic provided** (natural language):
- Use topic as exploration starting point
- Default to explore mode
- Load relevant memory bank context if topic relates to existing features/patterns

**Task ID provided** (e.g., TASK-042):
- Read task from `tasks.md`
- Load any existing plan or specification
- Default to refine mode for existing task, validate mode if task is "Ready for Implementation"

---

### Step 3: Load Relevant Context

**For all modes**, read:
- `activeContext.md` - Current project state
- `systemPatterns.md` - Established patterns and conventions
- `productContext.md` - Product requirements and features

**For specific task ID**, additionally read:
- Task entry from `tasks.md`
- Any related checkpoint entries from `progress.md`
- Planning output if task has been through /cf:plan

**For general topic**, search memory bank:
- Grep for topic-related keywords
- Load relevant sections of memory bank files

---

### Step 4: Engage Facilitator Agent

**Invoke Facilitator agent** from `.claude/agents/workflow/facilitator.md`.

**Provide context**:
- Loaded memory bank content
- Topic or task to facilitate
- Mode (explore/refine/validate)
- User's initial question or goal (if provided)

**Facilitator will**:
1. Analyze context for gaps, ambiguities, assumptions
2. Formulate clarifying questions organized by theme
3. Present questions to user for response
4. Listen to user's answers and ask follow-up questions
5. Iterate until topic is well-understood or plan is refined
6. Synthesize final understanding or refined specification
7. No iteration limit - user controls when to stop

---

### Step 5: Interactive Facilitation Loop

**Facilitator's question pattern**:

```markdown
🤝 FACILITATION SESSION: [Topic/Task Name]
─────────────────────────────

**Mode**: [Explore/Refine/Validate]
**Context**: [Brief context summary]

## Round 1: [Question Theme]

I'd like to understand [area] better. Here are some questions:

**[Category 1]**: [Category description]
1. [Clarifying question]
2. [Probing question]

**[Category 2]**: [Category description]
1. [Validating question]
2. [Edge case question]

**[Category 3]**: [Category description]
1. [Trade-off question]

---

Please answer the questions that are most relevant to you. We can iterate as needed.

**Type 'continue' when ready for synthesis, or answer specific questions to refine further.**

User provides answers:

• Can answer all questions, some questions, or add new information
• Can raise new questions or concerns
• Can request focus on specific area

Facilitator follow-up:

## Round 2: [Refined Theme Based on Answers]

Thank you for that context. Based on your answers, I have some follow-up questions:

**Clarification on [topic from user's answer]**:
1. You mentioned [X]. Does that mean [interpretation A] or [interpretation B]?
2. What happens if [edge case scenario]?

**New questions based on your answers**:
1. [Question arising from user's response]
2. [Question about implications]

---

We're making progress. [Summary of what's now clear]

**Still exploring**: [What remains ambiguous]

Continue answering, or type 'continue' if ready for synthesis.

Iteration continues until:

• User types 'continue' (ready for synthesis)
• User types 'stop' (end facilitation without synthesis)
• No more questions remain (full understanding achieved)

---

Step 6: Facilitator Synthesis

After user signals readiness, Facilitator produces synthesis:

Exploration Mode Synthesis

✨ EXPLORATION SYNTHESIS: [Topic]