AgentSkillsCN

task-breakdown

制定研究方法、信息搜集与成果呈现的标准,并妥善标注来源。在研究主题、分析文档,或综合多方信息时使用此功能。

SKILL.md
--- frontmatter
name: task-breakdown
description: Methodology for breaking specifications into actionable tasks organized by features. Use when creating task lists or orchestrating implementation.

Task Breakdown Skill

When to Use

  • Breaking specs into implementation tasks
  • Organizing work by features
  • Planning sequential or parallel execution
  • Estimating implementation effort

Feature Organization Methodology

Principle

UPMOST IMPORTANCE: The best path is the shortest to get working software that can be tested end-to-end.

Each feature delivers a complete user-facing capability that works across all application layers. Tasks are organized sequentially within each feature to achieve working, testable software as quickly as possible.

Feature Organization

Feature Structure:

code
## 1. [ ] Users should be able to signup

- 1.1 [ ] Create user migrations
- 1.2 [ ] Add /signup end-point
- 1.3 [ ] Build user signup form
- 1.4 [ ] Add user signup form validations
- 1.5 [ ] Send signup confirmation email
- 1.6 [ ] Add /signup/confirmation/:confirmation_id end-point
  - 1.6.1 [ ] Verify valid token
  - 1.6.2 [ ] Throw error with expired tokens
  - 1.6.3 [ ] Activate user account
- 1.7 [ ] Build user email confirmation page

## 2. [ ] Users should be able to login

- 2.1 [ ] Add /signin end-point
- 2.2 [ ] Build user sign-in form
- 2.3 [ ] Add user sign-in validations
- 2.4 [ ] Display "Lost password?" link after failed attempt

Feature Boundaries

Good feature: "Users should be able to signup"

  • Complete end-to-end functionality
  • Works across all layers (database, backend, frontend)
  • Can be tested and demonstrated
  • Delivers immediate user value

Bad feature: "All database migrations"

  • Horizontal layer, not vertical feature
  • Cannot be tested end-to-end
  • Doesn't deliver user value by itself

Task Types Within Features

As you break down each feature, you'll typically encounter these types of tasks in logical sequence:

  • Data layer: Migrations, schema design, indexes
  • Backend: API endpoints, business logic, validation, error handling
  • Frontend: UI components, forms, user flows, state management
  • Integration: Email sending, external APIs, webhooks
  • Testing: Unit tests, integration tests, E2E tests

The key is to organize these in the shortest path to working software, not in separate sections.

Estimation Guidelines

Complexity Levels:

  • Simple (1 point): Straightforward, well-understood, minimal dependencies
  • Medium (3 points): Moderate complexity, some unknowns, few dependencies
  • Complex (5 points): High complexity, many unknowns, multiple dependencies
  • Very Complex (8 points): Requires research, high risk, many dependencies

Red Flags:

  • Tasks larger than 8 points (break down further)
  • Vague task descriptions
  • Hidden dependencies

Dependency Analysis

  1. Identify Dependencies:

    • What must be done before this task?
    • What blocks this task?
    • What does this task enable?
  2. Critical Path:

    • Longest chain of dependent tasks
    • High-risk items should be tackled early
  3. Parallel Opportunities:

    • Tasks with no dependencies can run in parallel
    • Different specialties can work simultaneously within a feature

Orchestration Strategies

Sequential Execution

  • One feature at a time, in priority order
  • Complete all tasks for feature 1, then move to feature 2
  • Each feature delivers working, testable software before moving to the next
  • Best for: Learning codebases, tight coordination needed, high-risk features

Parallel Execution

  • Multiple features simultaneously
  • Multiple subagents working on independent features
  • Best for: Large specs with independent features, experienced teams, well-defined boundaries
  • Requires careful dependency management

Formatting Standards

CRITICAL: Follow template formatting exactly:

  • Use text checkboxes: [ ] for incomplete, [x] for complete
  • NEVER use emojis for checkboxes or task status
  • ALL tasks MUST have task IDs in the format: 1.1, 1.2, 2.1, 2.2, etc.
  • ALL subtasks MUST have nested IDs like: 1.3.1, 1.3.2, 1.3.3
  • Follow the exact structure and styling from tasks-template.md
  • Maintain consistent indentation (2 spaces for nested items)
  • Feature headers use: ## N. [ ] Feature description
  • Tasks use: - N.M [ ] Task description
  • Subtasks use: - N.M.K [ ] Subtask description

Templates

See templates/ directory for:

  • tasks-template.md