task-o-matic（任务自动化）

Task-O-Matic

AI-powered task management CLI for single projects. Helps initialize projects, parse PRDs into tasks, break down work with AI, and execute tasks with automatic retry and verification.

When to Use

Use when the user wants to:

• Initialize a new project with task-o-matic
• Attach task-o-matic to an existing project (detect stack automatically)
• Parse a PRD (Product Requirements Document) into structured tasks
• Break down tasks into smaller subtasks using AI
• Enhance tasks with AI-powered documentation research
• Execute tasks with AI coding assistants
• Configure AI providers and models

CLI Location

The CLI binary is at: packages/cli/dist/cli/bin.js (after building)

Run with: node packages/cli/dist/cli/bin.js <command> [options]

Or if installed globally: task-o-matic <command> [options]

Core Workflow

# New project workflow
1. Initialize project (task-o-matic init init)
2. Configure AI provider (task-o-matic config set-ai-provider)
3. Parse PRD into tasks (task-o-matic prd parse)
4. Split tasks into subtasks (task-o-matic tasks split --all)
5. Execute tasks (task-o-matic tasks execute-loop)

# Existing project workflow
1. Attach to existing project (task-o-matic init attach)
2. Configure AI provider (task-o-matic config set-ai-provider)
3. Generate PRD from codebase (task-o-matic prd generate --from-codebase)
4. Create tasks from PRD (task-o-matic prd parse)
5. Execute tasks (task-o-matic tasks execute-loop)

Common Commands

Initialize Project

# Basic initialization
task-o-matic init init --project-name my-app --package-manager bun

# With specific AI provider
task-o-matic init init --project-name my-app --ai-provider openrouter --ai-model xiaomi/mimo-v2-flash:free

# With Better-T-Stack bootstrap
task-o-matic init init --project-name my-app --frontend next --backend convex --auth better-auth

Options:

• --ai-provider <provider>: AI provider (openrouter/anthropic/openai/custom)
• --ai-model <model>: AI model
• --project-name <name>: Project name
• --package-manager <pm>: bun/npm/pnpm (default: npm)
• --frontend <frontends...>: Frontend framework(s) (default: next)
• --backend <backend>: Backend framework (default: convex)
• --auth <auth>: Authentication (default: better-auth)
• --directory <dir>: Working directory

Attach to Existing Project

# Attach to existing project (auto-detect stack)
task-o-matic init attach

# With full project analysis
task-o-matic init attach --analyze

# Dry run (just show detection, don't create files)
task-o-matic init attach --dry-run

# Force re-detection (update cached stack.json)
task-o-matic init attach --redetect

Options:

• --analyze: Run full project analysis (TODOs, features, structure)
• --dry-run: Just detect stack, don't create files
• --redetect: Force re-detection (overwrites cached stack.json)
• --ai-provider <provider>: AI provider (openrouter/anthropic/openai/custom)
• --ai-model <model>: AI model
• --context7-api-key <key>: Context7 API key

What it detects:

• Language (TypeScript/JavaScript)
• Framework(s) (Next.js, Express, Hono, etc.)
• Database (Postgres, MongoDB, SQLite, etc.)
• ORM (Prisma, Drizzle, etc.)
• Auth (Better-Auth, Clerk, NextAuth, etc.)
• Package Manager (npm, pnpm, bun, yarn)
• Runtime (Node, Bun)
• API style (tRPC, GraphQL, REST)
• Testing frameworks
• Build tools

What it creates:

• .task-o-matic/config.json - AI settings
• .task-o-matic/stack.json - Cached stack detection (used by AI context)
• .task-o-matic/mcp.json - Context7 config
• .task-o-matic/tasks/ - Task storage
• .task-o-matic/prd/ - PRD storage
• Updates .gitignore if git exists

Configure AI Provider

# Set provider and model
task-o-matic config set-ai-provider openrouter xiaomi/mimo-v2-flash:free

# Set with custom URL
task-o-matic config set-ai-provider custom https://api.example.com

# Check current config
task-o-matic config get-ai-config

Generate PRD from Codebase

# Analyze current project and generate PRD
task-o-matic prd generate

# With streaming output
task-o-matic prd generate --stream

# With custom output filename
task-o-matic prd generate --output my-project-prd.md

Parse PRD

# Basic parse
task-o-matic prd parse --file prd.md

# With streaming output
task-o-matic prd parse --file prd.md --stream

# With reasoning tokens (for models that support it)
task-o-matic prd parse --file prd.md --ai-reasoning 4000 --stream

# Multiple models with combination
task-o-matic prd parse --file prd.md --ai openrouter:model1 openrouter:model2 --combine-ai openrouter:combine-model

Options:

• --file <path>: Path to PRD file (required)
• --stream: Show streaming AI output
• --ai-reasoning <tokens>: Max reasoning tokens for supported models
• --ai <models...>: Multiple AI models
• --combine-ai <provider:model>: Model to combine results
• --tools: Enable filesystem tools for project analysis

Work with Tasks

List Tasks

task-o-matic tasks list

# With filters
task-o-matic tasks list --status todo
task-o-matic tasks list --tag frontend

Create Task

# Manual creation
task-o-matic tasks create --title "Add authentication" --content "Implement OAuth2 login"

# With AI enhancement
task-o-matic tasks create --title "Add authentication" --ai-enhance --stream

Split Tasks

# Split specific task
task-o-matic tasks split --task-id 1 --stream

# Split all tasks
task-o-matic tasks split --all --stream

# With filters
task-o-matic tasks split --all --status todo --stream

Options:

• --task-id <id>: Specific task to split
• --all: Split all tasks without subtasks
• --status <status>: Filter by status (todo/in-progress/completed)
• --stream: Show streaming output
• --ai-reasoning <tokens>: Enable reasoning
• --dry: Preview without changes

Enhance Tasks

# Enhance specific task
task-o-matic tasks enhance --task-id 1 --stream

# Enhance all tasks
task-o-matic tasks enhance --all --stream

# With filter
task-o-matic tasks enhance --all --status todo --stream

Options:

• --task-id <id>: Specific task to enhance
• --all: Enhance all tasks
• --status <status>: Filter by status
• --tag <tag>: Filter by tag
• --dry: Preview without changes
• --force: Skip confirmation

Task Status

# Set task status
task-o-matic tasks status --task-id 1 --status in-progress

# Available statuses: todo, in-progress, completed

Task Tree

# Show hierarchical task tree
task-o-matic tasks tree

Get Next Task

# Get next task to work on
task-o-matic tasks get-next

Execute Tasks

# Execute all todo tasks
task-o-matic tasks execute-loop --status todo

# Execute specific tasks
task-o-matic tasks execute-loop --ids 1,2,3

# With specific tool
task-o-matic tasks execute-loop --tool claude

# With planning phase
task-o-matic tasks execute-loop --plan --review-plan

# With verification
task-o-matic tasks execute-loop --verify "bun run test" --verify "bun run type-check"

# Progressive model escalation
task-o-matic tasks execute-loop --try-models "gpt-4o-mini,gpt-4o,claude:sonnet-4"

Options:

• --status <status>: Filter by status
• --tag <tag>: Filter by tag
• --ids <ids>: Comma-separated task IDs
• --tool <tool>: opencode/claude/gemini/codex (default: opencode)
• --max-retries <number>: Max retries per task (default: 3)
• --model <model>: Force specific model
• --verify <command>: Verification command (can be used multiple times)
• --plan: Generate implementation plan before execution
• --plan-model <model>: Model for planning
• --review-plan: Pause for human review of plan
• --review: Run AI review after execution
• --auto-commit: Auto-commit after each task
• --try-models <models>: Progressive models for retries
• --dry: Show what would execute

PRD Management

Create PRD from Description

task-o-matic prd create "Build a task management app with real-time updates"

Refine PRD with Questions

# Generate questions, answer interactively, refine PRD
task-o-matic prd refine --file prd.md --output prd_refined.md --stream

# Questions mode: AI generates clarifying questions
task-o-matic prd question --file prd.md

Combine PRDs

task-o-matic prd combine --files prd1.md,prd2.md --output master.md

Get Tech Stack Suggestion

task-o-matic prd get-stack --file prd.md

Interactive Workflow

The workflow command provides an all-in-one interactive experience:

# Full interactive workflow
task-o-matic workflow

# Skip specific steps
task-o-matic workflow --skip-bootstrap --skip-prd-question-refine

# Auto-accept all suggestions
task-o-matic workflow --auto-accept

# Use config file
task-o-matic workflow --config-file workflow-config.json

Workflow covers:

• Project initialization
• PRD creation/upload
• PRD question & refinement
• Task generation
• Task splitting
• Task execution

AI Providers

OpenRouter

task-o-matic config set-ai-provider openrouter xiaomi/mimo-v2-flash:free

Popular free models:

• xiaomi/mimo-v2-flash:free
• google/gemma-3-27b-it:free
• meta-llama/llama-3.3-8b-instruct:free

Anthropic

task-o-matic config set-ai-provider anthropic claude-3-5-sonnet-20241022

OpenAI

task-o-matic config set-ai-provider openai gpt-4o

Custom

task-o-matic config set-ai-provider custom https://api.example.com

Environment Variables

Set in .env file or environment:

AI_PROVIDER=openrouter
AI_MODEL=xiaomi/mimo-v2-flash:free
OPENROUTER_API_KEY=your_key_here
ANTHROPIC_API_KEY=your_key_here
OPENAI_API_KEY=your_key_here

Project Structure

After initialization:

.task-o-matic/
├── config.json       # Project configuration
├── tasks.json        # Tasks database
├── prd.md            # PRD file
└── plans/            # Implementation plans

Best Practices

1. Always use --stream for long-running AI operations to see progress
2. Use --ai-reasoning with models that support it for better task breakdown
3. Filter by status when splitting/enhancing to avoid reprocessing completed tasks
4. Use --dry first for bulk operations to preview changes
5. Start with free models (OpenRouter free tier) before using paid models
6. Use --verify during execution to run tests after each task
7. Enable --plan for complex tasks to get implementation plans first

Example Session

# 1. Initialize
task-o-matic init init --project-name my-app --package-manager bun

# 2. Configure AI (optional if set in .env)
task-o-matic config set-ai-provider openrouter xiaomi/mimo-v2-flash:free

# 3. Parse PRD
task-o-matic prd parse --file prd.md --stream --ai-reasoning 4000

# 4. Review tasks
task-o-matic tasks tree

# 5. Split all tasks
task-o-matic tasks split --all --stream --ai-reasoning 4000

# 6. Enhance tasks with documentation
task-o-matic tasks enhance --all --stream

# 7. Execute
task-o-matic tasks execute-loop --status todo --verify "bun run test" --plan

Advanced Topics

Progressive Model Escalation

Use cheaper models first, escalate to better ones on failure:

task-o-matic tasks execute-loop \
  --try-models "xiaomi/mimo-v2-flash:free,gpt-4o-mini,claude:sonnet-4" \
  --max-retries 3

Planning Phase

Generate implementation plans before execution:

task-o-matic tasks execute-loop \
  --plan \
  --plan-model gpt-4o \
  --review-plan

Custom Verification

Run multiple verification commands:

task-o-matic tasks execute-loop \
  --verify "bun run type-check" \
  --verify "bun run test" \
  --verify "bun run lint"

See Also

• Complete Command Reference - Detailed command options
• Project README - Full project documentation
• Tutorial - Step-by-step guide