---
name: copilot-plan
description: "Parallel planning: run a second AI model alongside Claude Code to generate competing implementation plans. Compare and pick the best approach."
---

# Parallel AI Planning — Competing Plans Comparison

Generate implementation plans from two AI models in parallel, then compare and combine the best ideas.

## When to Use

- When planning a new feature or significant change
- When you want to compare approaches from different models
- After brainstorming, as an alternative to going straight to a single plan

## How It Works

1. Takes a feature description or design doc
2. Launches a second AI (e.g., GitHub Copilot CLI, OpenAI API, or another tool) in the background
3. Meanwhile, Claude Code generates its own plan
4. Presents both plans side by side for comparison

## Process

### Step 1: Gather the Brief

Ask the user for:
- **What to build** — feature description, design doc path, or issue number
- **Project context** — which repo/directory, tech stack
- **Constraints** — any must-haves or limitations
- **Second model** — which tool to use (default: GitHub Copilot CLI)

### Step 2: Launch Second Model Planning in Background

For GitHub Copilot CLI:
```bash
# Adjust this command for your second model tool
echo "<brief>" | gh copilot suggest -t shell "Create an implementation plan for: <brief>" > /tmp/copilot-plan.md &
```

For other tools, adapt the command accordingly (OpenAI CLI, local model, etc.).

### Step 3: Generate Claude Code Plan

While the second model runs, create your own implementation plan:
- Bite-sized tasks (2-5 min each)
- Exact file paths and complete code
- TDD approach where applicable
- DRY, YAGNI

### Step 4: Present Comparison

Once both are complete:

```markdown
## Plan A: [Second Model Name]
[Their plan]

## Plan B: Claude Code
[Your plan]

## Comparison
| Aspect | [Second Model] | Claude Code |
|--------|---------------|-------------|
| Architecture | ... | ... |
| Task breakdown | ... | ... |
| Testing strategy | ... | ... |
| Estimated steps | ... | ... |
| Unique insights | ... | ... |

## Recommendation
[Which plan or which elements from each to combine]
```

### Step 5: User Picks

The user chooses Plan A, Plan B, or a hybrid. Save the chosen plan and proceed to implementation.

## Supported Second Models

| Tool | Command | Notes |
|------|---------|-------|
| GitHub Copilot CLI | `gh copilot suggest` | Requires Copilot subscription |
| OpenAI CLI | `openai api chat.completions.create` | Requires OpenAI API key |
| Local model (Ollama) | `ollama run <model>` | Free, runs locally |
| Any CLI tool | Adapt the background command | Any tool that accepts text input |

## Setup

1. Install your preferred second model CLI tool
2. Configure authentication (API keys, tokens, etc.)
3. Update the background command in Step 2 to match your tool