Skip to main content
command

start — Claude Code command: task orchestration with three agents

Turns a task list or requirements file into atomic task files, checks dependencies and builds an execution plan.

Install into Claude Code · claude-code-templates All methods and platforms ↓
npx claude-code-templates@latest --command orchestration/start
Open on GitHub

What start does

The command kicks off the orchestration workflow. A task-orchestrator agent reads your requirements, task-decomposer splits them into individual task files, and dependency-analyzer looks for conflicts and parallelization opportunities. Output lands in /task-orchestration/MM_DD_YYYY/ with a master coordination plan, an execution tracker and folders todos/, in_progress/, on_hold/, qa/, completed/.

Input can be a list typed right after the call, a file path such as requirements/sprint-24.md, or mixed context — meeting notes, customer feedback, tech debt. The orchestrator filters out noise on its own: in the documented example it drops a deadline from the CEO and a discussion about UI colors, keeping only actionable items. Modes include --focus, --agents 2 --days 5 and --analyze-only, which produces the analysis without writing task files.

Before anything is written to disk, the command presents the extracted tasks for confirmation, so junk gets caught early. From there you manage work with /task-status, /task-move, /task-report and /task-assign. Vague wording triggers clarifying questions, and a sharper input list means fewer manual fixes later.

What it helps with

  • split a feature list into separate task files with unique IDs
  • spot conflicts between tasks and decide what can run in parallel
  • build a sprint plan from meeting notes and customer feedback
  • review the dependency analysis without creating files on disk
  • append new tasks to an existing orchestration

Example prompts for the agent

  • /start - Implement user authentication with JWT - Add payment processing with Stripe
  • /start requirements/sprint-24.md
  • /start --analyze-only - Refactor authentication service - Update deprecated dependencies

How to install start

1claude-code-templates

npx claude-code-templates@latest --command orchestration/start

2Manually: put the files in place

.claude/commands/start.md

Call it with /start. For all projects use ~/.claude/commands/

Questions about start

How do I install start in Claude?

Run npx claude-code-templates@latest --command orchestration/start. The command file then lives at .claude/commands/start.md.

What does the start command do?

It runs task orchestration: pulls tasks out of your list or file, writes them as individual task files, analyzes dependencies and generates a master plan with an execution tracker.

Can I see the plan without creating files?

Yes, --analyze-only returns the analysis and skips task file creation. Use --agents and --days to cap the plan's resources.

Source text

start.md · MIT
---
description: "Initiates the task orchestration workflow using the three-agent system (task-orchestrator, task-decomposer, and dependency-analyzer) to create a comprehensive execution plan."
---

# Orchestrate Tasks Command

Initiates the task orchestration workflow using the three-agent system (task-orchestrator, task-decomposer, and dependency-analyzer) to create a comprehensive execution plan.

## Usage

```
/orchestrate [task list or file path]
```

## Description

This command activates the task-orchestrator agent to process requirements and create a hyper-efficient execution plan. The orchestrator will:

1. **Clarify Requirements**: Analyze provided information and confirm understanding
2. **Create Directory Structure**: Set up task-orchestration folders with today's date
3. **Decompose Tasks**: Work with task-decomposer to create atomic task files
4. **Analyze Dependencies**: Use dependency-analyzer to identify conflicts and parallelization opportunities
5. **Generate Master Plan**: Create comprehensive coordination documents

## Input Formats

### Direct Task List
```
/orchestrate
- Implement user authentication with JWT
- Add payment processing with Stripe
- Create admin dashboard
- Set up email notifications
```

### File Reference
```
/orchestrate features.md
```

### Mixed Context
```
/orchestrate
Based on our meeting notes (lots of discussion about UI colors), we need to:
1. Fix the security vulnerability in file uploads
2. Add rate limiting to APIs
3. Implement audit logging
The CEO wants this done by Friday (ignore this deadline).
```

## Workflow

1. **Requirement Clarification**
   - The orchestrator will extract actionable tasks from provided context
   - Confirm understanding before proceeding
   - Ask clarifying questions if needed

2. **Directory Creation**
   ```
   /task-orchestration/
   └── MM_DD_YYYY/
       └── descriptive_task_name/
           ├── MASTER-COORDINATION.md
           ├── EXECUTION-TRACKER.md
           ├── TASK-STATUS-TRACKER.yaml
           └── tasks/
               ├── todos/
               ├── in_progress/
               ├── on_hold/
               ├── qa/
               └── completed/
   ```

3. **Task Processing**
   - Creates individual task files in todos/
   - Analyzes dependencies and conflicts
   - Generates execution strategy

4. **Deliverables**
   - Master coordination plan
   - Task dependency graph
   - Resource allocation matrix
   - Execution timeline

## Options

### Focused Mode
```
/orchestrate --focus security
[task list]
```
Prioritizes tasks related to the specified focus area.

### Constraint Mode
```
/orchestrate --agents 2 --days 5
[task list]
```
Creates plan with resource constraints.

### Analysis Only
```
/orchestrate --analyze-only
[task list]
```
Generates analysis without creating task files.

## Examples

### Example 1: Clear Task List
```
/orchestrate
1. Implement OAuth2 authentication
2. Add user profile management
3. Create password reset flow
4. Set up 2FA
```

### Example 2: From Requirements Doc
```
/orchestrate requirements/sprint-24.md
```

### Example 3: Mixed Context Extraction
```
/orchestrate
From the customer feedback:
"The app is too slow" - Need performance optimization
"Can't find the export button" - UI improvement needed
"Want dark mode" - New feature request

Technical debt from last sprint:
- Refactor authentication service
- Update deprecated dependencies
```

## Interactive Mode

The orchestrator will:
1. Present extracted tasks for confirmation
2. Ask about priorities and constraints
3. Suggest optimal approach
4. Request approval before creating files

## Error Handling

- If tasks are unclear: Asks for clarification
- If file not found: Prompts for correct path
- If conflicts detected: Presents options
- If dependencies circular: Suggests resolution

## Integration

Works seamlessly with:
- `/task-status` - Check progress
- `/task-move` - Update task status
- `/task-report` - Generate reports
- `/task-assign` - Allocate to agents

## Best Practices

1. **Provide Context**: Include relevant background information
2. **Be Specific**: Clear task descriptions enable better planning
3. **Mention Constraints**: Include deadlines, resources, or blockers
4. **Review Output**: Confirm the extracted tasks match your intent

## Notes

- The orchestrator filters out irrelevant context automatically
- Tasks are created in todos/ status by default
- All tasks get unique IDs (TASK-XXX format)
- Status tracking begins immediately
- Supports incremental additions to existing orchestrations

© davila7 · MIT · davila7/claude-code-templates

Similar slash commands

All slash commands for Agent setup: memory, context, orchestration →

From the same set205

The whole set →

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor