mirror-magi-meta-agent
Version:
AI-powered development planning and execution system with Supabase integration
233 lines (185 loc) • 5.48 kB
Markdown
# ✅ AI Plan Validation & Execution
*Streamlined workflow: AI generates → You structure → Validate → Execute*
## 🎯 Simplified Workflow
### Your Process:
1. **Generate plan with AI** (ChatGPT, Claude, etc.)
2. **Structure it externally** (copy/paste into JSON format)
3. **Validate with our tool** (ensure exact correct structure)
4. **Execute step-by-step** (get Claude Code Max commands)
## ✅ Plan Validation
### Quick Validation Command
```bash
npm run plan:validate
```
### Options:
1. **Validate new plan** - Paste your JSON plan for validation
2. **Validate existing plan** - Check saved plan file
3. **Show structure requirements** - See exact JSON format needed
## 📋 Required Structure
Your AI-generated plan must have this exact JSON structure:
```json
{
"id": "plan_1703089234567",
"goal": "Clear project goal statement",
"status": "ready",
"created": "2024-12-20T14:30:00.000Z",
"phases": [
{
"id": "phase_1703089234567_1",
"name": "Phase Name",
"description": "What this phase accomplishes",
"tasks": [
{
"id": "task_1703089234567_001",
"description": "Specific task description",
"type": "component_creation",
"dependencies": [],
"specifics": {
"componentName": "MyComponent",
"filePath": "src/components/MyComponent.tsx"
}
}
]
}
]
}
```
## ✅ Required Fields Checklist
**Plan Level:**
- ✅ `id` - Unique plan identifier
- ✅ `goal` - Clear project goal
- ✅ `phases` - Array of phases
**Phase Level:**
- ✅ `id` - Unique phase identifier
- ✅ `name` - Phase name
- ✅ `description` - Phase description
- ✅ `tasks` - Array of tasks
**Task Level:**
- ✅ `id` - Unique task identifier
- ✅ `description` - Task description
- ✅ `type` - Task type (see valid types below)
- ✅ `dependencies` - Array of task IDs (can be empty: `[]`)
- ✅ `specifics` - Object with details (can be empty: `{}`)
## 🏷️ Valid Task Types
- `component_creation` - Building UI components
- `api_integration` - API connections and data fetching
- `configuration` - Config files, routing, build setup
- `testing_implementation` - Tests and validation
- `feature_development` - General functionality
- `styling_implementation` - CSS and design
- `data_modeling` - Types and data structures
## 🔍 Validation Process
### Step 1: Run Validator
```bash
npm run plan:validate
```
### Step 2: Choose Option 1
- Paste your JSON plan
- End with line containing just "END"
### Step 3: Review Results
- ✅ **VALID**: Plan ready for execution
- ❌ **ERRORS**: Must fix before execution
- ⚠️ **WARNINGS**: Recommended improvements
### Step 4: Save & Execute
If valid:
- Plan saves to `state/master-plan.json`
- Ready for execution with `npm run plan:continue`
## 🚀 Execution Commands
After validation passes:
```bash
npm run plan:continue # Get first Claude Code Max command
# → Copy command to Claude Code Max
# → Let Claude implement the task
node plan-project.js complete [task_id] # Mark task complete
npm run plan:continue # Get next command
# → Repeat until all tasks done
```
## 📊 Other Useful Commands
```bash
npm run plan:view # See complete plan with all details
npm run plan:progress # Check completion status
npm run plan:edit # Make manual edits (optional)
```
## ⚠️ Common Validation Errors
### Missing Required Fields
```json
// ❌ Missing required fields
{
"description": "Build something"
// Missing: id, type, dependencies, specifics
}
// ✅ All required fields present
{
"id": "task_123_001",
"description": "Build something",
"type": "component_creation",
"dependencies": [],
"specifics": {}
}
```
### Invalid Dependencies
```json
// ❌ Reference to non-existent task
{
"dependencies": ["task_does_not_exist"]
}
// ✅ Valid dependency reference
{
"dependencies": ["task_123_001"] // This task must exist in plan
}
```
### Duplicate IDs
```json
// ❌ Same ID used twice
{
"tasks": [
{"id": "task_1", "description": "Task A"},
{"id": "task_1", "description": "Task B"} // Duplicate!
]
}
```
## 💡 Pro Tips
### Better Structure:
- Use consistent ID patterns: `task_timestamp_001`
- Keep task descriptions specific and actionable
- Group related tasks in logical phases
- Ensure dependencies make chronological sense
### Task Specifics:
- Add file paths for component tasks
- Include API endpoints for integration tasks
- Specify config files for configuration tasks
- The more specific, the better Claude Code Max commands
### Validation First:
- Always validate before execution
- Fix all errors (not just warnings)
- Check that first task preview makes sense
- Ensure total task count seems reasonable
## 🎯 Example Valid Task
```json
{
"id": "task_1703089234567_001",
"description": "Create UserDashboard component with responsive layout and analytics props",
"type": "component_creation",
"dependencies": [],
"specifics": {
"componentName": "UserDashboard",
"filePath": "src/components/UserDashboard.tsx",
"props": {
"user": "User",
"analytics": "AnalyticsData"
}
}
}
```
This will generate a specific Claude Code Max command like:
> "Create a UserDashboard component in src/components/UserDashboard.tsx with props for user and analytics..."
*Perfect structure = Perfect execution commands*