Claude Code Plugins

Community-maintained marketplace

Feedback

Breaks down specs into complexity-stratified tasks. Use when organizing tasks, assigning work to agents, or when user mentions task layering, task stratification, or parallel task execution.

Install Skill

1Download skill
2Enable skills in Claude

Open claude.ai/settings/capabilities and find the "Skills" section

3Upload to Claude

Click "Upload skill" and select the downloaded ZIP file

Note: Please verify skill by going through its instructions before using it.

SKILL.md

name Task Layering
description Breaks down specs into complexity-stratified tasks. Use when organizing tasks, assigning work to agents, or when user mentions task layering, task stratification, or parallel task execution.
allowed-tools Read, Write, Bash

Task Layering

This skill provides task stratification by complexity and agent assignment for parallel execution.

What This Skill Provides

1. Complexity Stratification

  • Trivial (0): < 5 minutes, simple changes
  • Simple (1): < 30 minutes, straightforward implementation
  • Moderate (2): < 2 hours, requires thought
  • Complex (3): > 2 hours, architectural decisions

2. Agent Assignment Strategy

  • @claude: Complex (3), security, architecture
  • @copilot: Trivial (0), Simple (1)
  • @qwen: Performance optimization
  • @gemini: Documentation
  • @codex: Testing, TDD

3. Layer Organization

  • Foundation Layer: Must complete first (dependencies)
  • Parallel Layer: Can execute simultaneously
  • Integration Layer: Combines parallel work

4. Output Files

  • layered-tasks.md - All tasks stratified by complexity
  • claude-tasks.md - Tasks assigned to Claude
  • copilot-tasks.md - Tasks assigned to Copilot
  • qwen-tasks.md - Tasks for performance work
  • gemini-tasks.md - Documentation tasks
  • codex-tasks.md - Testing tasks

Instructions

Stratifying Tasks

When user wants to layer tasks from a spec:

  1. Read the spec's tasks.md file

  2. Analyze each task for:

    • Estimated time to complete
    • Required expertise level
    • Dependencies on other tasks
    • Appropriate agent assignment
  3. Create layered-tasks.md with format:

    # Layered Tasks: [Spec Name]
    
    ## Foundation Layer (Sequential)
    - [Complex] Task 1 (@claude) - Requires arch decisions
    - [Moderate] Task 2 (@claude) - Core implementation
    
    ## Parallel Layer (Concurrent)
    - [Simple] Task 3 (@copilot) - UI component
    - [Simple] Task 4 (@copilot) - Data model
    - [Moderate] Task 5 (@qwen) - Query optimization
    
    ## Integration Layer (Sequential)
    - [Moderate] Task 6 (@claude) - Integrate components
    - [Simple] Task 7 (@codex) - Integration tests
    
  4. Create agent-specific task files for distribution

Task Complexity Analysis

Trivial (0) Examples:

  • Update constant value
  • Fix typo
  • Add simple validation

Simple (1) Examples:

  • Create CRUD component
  • Add new API endpoint
  • Write unit test

Moderate (2) Examples:

  • Implement authentication flow
  • Optimize database queries
  • Refactor module structure

Complex (3) Examples:

  • Design system architecture
  • Implement security layer
  • Create caching strategy

Layered Tasks Format

# Layered Tasks: Feature Name

## Summary
- Total tasks: 15
- Foundation: 3 tasks
- Parallel: 10 tasks
- Integration: 2 tasks

## Foundation Layer
Must complete in order before parallel work:

### 1. [Complex] Design Data Schema (@claude)
**Dependencies**: None
**Estimate**: 3 hours
**Description**: Design database schema and relationships

## Parallel Layer
Can execute simultaneously:

### 2. [Simple] Create User Model (@copilot)
**Dependencies**: Task 1
**Estimate**: 20 minutes

### 3. [Simple] Create Auth Component (@copilot)
**Dependencies**: Task 1
**Estimate**: 30 minutes

## Integration Layer
Combine parallel work:

### 15. [Moderate] Integration Testing (@codex)
**Dependencies**: All parallel tasks
**Estimate**: 1 hour

Success Criteria

  • ✅ Tasks are stratified by realistic complexity
  • ✅ Dependencies are identified correctly
  • ✅ Agent assignments match expertise
  • ✅ Layers enable maximum parallelization
  • ✅ Output files are complete and actionable

Plugin: 04-iterate Skill Type: Analysis + Organization Auto-invocation: Yes (via description matching)