agentsclimarketplace

Planning and task breakdown

Skill vinvcn/addyosmani-agent-skills-zh/skills/planning-and-task-breakdown

将工作拆解为有序任务。用于已有 spec 或清晰需求,并需要拆解成可实现任务时。用于任务太大难以下手、需要估算范围,或存在并行工作机会时。From its SKILL.md

Install
npx -y skills add vinvcn/addyosmani-agent-skills-zh --skill planning-and-task-breakdown

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 23 stars23 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

6.4 KB, ~2.1k tokens by cl100k_base, as published. Nobody here has run it

Planning and Task Breakdown

概览

把工作拆解成小而可验证的任务,并为每个任务写明验收标准。好的任务拆解,是可靠完成工作的 agent 和产出纠缠混乱结果的 agent 之间的区别。每个任务都应该足够小,能在一次专注会话中实现、测试和验证。

何时使用

  • 你已有 spec,需要把它拆成可实现单元
  • 一个任务感觉太大或太模糊,难以下手
  • 工作需要跨多个 agents 或会话并行
  • 你需要向人类沟通范围
  • 实现顺序并不明显

何时不要使用: 范围显而易见的单文件变更,或 spec 已经包含定义良好的任务。

计划流程

步骤 1:进入 Plan Mode

写任何代码之前,以只读模式运行:

  • 阅读 spec 和相关代码库区域
  • 识别现有模式和约定
  • 映射组件之间的依赖
  • 记录风险和未知项

规划期间不要写代码。 输出是一份计划文档,而不是实现。

步骤 2:识别依赖图

映射什么依赖什么:

Database schema
    │
    ├── API models/types
    │       │
    │       ├── API endpoints
    │       │       │
    │       │       └── Frontend API client
    │       │               │
    │       │               └── UI components
    │       │
    │       └── Validation logic
    │
    └── Seed data / migrations

实现顺序沿依赖图自底向上:先构建基础。

步骤 3:垂直切片

不要先构建全部数据库、再构建全部 API、再构建全部 UI。一次构建一条完整功能路径:

坏例(水平切片):

Task 1: Build entire database schema
Task 2: Build all API endpoints
Task 3: Build all UI components
Task 4: Connect everything

好例(垂直切片):

Task 1: User can create an account (schema + API + UI for registration)
Task 2: User can log in (auth schema + API + UI for login)
Task 3: User can create a task (task schema + API + UI for creation)
Task 4: User can view task list (query + API + UI for list view)

每个垂直切片都交付可工作的、可测试的功能。

步骤 4:编写任务

每个任务遵循这个结构:

## Task [N]: [Short descriptive title]

**Description:** One paragraph explaining what this task accomplishes.

**Acceptance criteria:**
- [ ] [Specific, testable condition]
- [ ] [Specific, testable condition]

**Verification:**
- [ ] Tests pass: `npm test -- --grep "feature-name"`
- [ ] Build succeeds: `npm run build`
- [ ] Manual check: [description of what to verify]

**Dependencies:** [Task numbers this depends on, or "None"]

**Files likely touched:**
- `src/path/to/file.ts`
- `tests/path/to/test.ts`

**Estimated scope:** [Small: 1-2 files | Medium: 3-5 files | Large: 5+ files]

步骤 5:排序并设置检查点

安排任务时确保:

  1. 依赖已满足(先构建基础)
  2. 每个任务结束时系统仍处于可工作状态
  3. 每 2-3 个任务后出现验证检查点
  4. 高风险任务靠前(快速失败)

添加明确检查点:

## Checkpoint: After Tasks 1-3
- [ ] All tests pass
- [ ] Application builds without errors
- [ ] Core user flow works end-to-end
- [ ] Review with human before proceeding

任务大小指南

大小文件数范围示例
XS1单个函数或配置变更添加一条验证规则
S1-2一个组件或 endpoint添加一个新 API endpoint
M3-5一个功能切片用户注册流程
L5-8多组件功能带过滤和分页的搜索
XL8+太大,需要进一步拆解

如果任务是 L 或更大,就应该拆成更小任务。agent 在 S 和 M 任务上表现最好。

何时进一步拆分任务:

  • 它会超过一次专注会话(约 2+ 小时 agent 工作)
  • 你无法用 3 条或更少 bullet 描述验收标准
  • 它触及两个或更多独立子系统(例如 auth 和 billing)
  • 你发现自己在任务标题里写 “and”(这通常说明它是两个任务)

计划文档模板

# Implementation Plan: [Feature/Project Name]

## Overview
[One paragraph summary of what we're building]

## Architecture Decisions
- [Key decision 1 and rationale]
- [Key decision 2 and rationale]

## Task List

### Phase 1: Foundation
- [ ] Task 1: ...
- [ ] Task 2: ...

### Checkpoint: Foundation
- [ ] Tests pass, builds clean

### Phase 2: Core Features
- [ ] Task 3: ...
- [ ] Task 4: ...

### Checkpoint: Core Features
- [ ] End-to-end flow works

### Phase 3: Polish
- [ ] Task 5: ...
- [ ] Task 6: ...

### Checkpoint: Complete
- [ ] All acceptance criteria met
- [ ] Ready for review

## Risks and Mitigations
| Risk | Impact | Mitigation |
|------|--------|------------|
| [Risk] | [High/Med/Low] | [Strategy] |

## Open Questions
- [Question needing human input]

并行机会

当多个 agents 或会话可用时:

  • 可以安全并行: 独立功能切片、已实现功能的测试、文档
  • 必须顺序执行: 数据库迁移、共享状态变更、依赖链
  • 需要协调: 共享 API 契约的功能(先定义契约,再并行)

常见合理化借口

合理化借口现实
“我边做边想”这会让你得到纠缠混乱和返工。10 分钟规划能节省数小时。
“任务很明显”无论如何都写下来。显式任务会暴露隐藏依赖和被遗忘的边界情况。
“规划是额外开销”规划就是任务。没有计划的实现只是打字。
“我可以全都记在脑子里”上下文窗口是有限的。书面计划能跨会话边界和压缩继续存在。

红旗

  • 没有书面任务列表就开始实现
  • 任务只写“实现功能”,没有验收标准
  • 计划中没有验证步骤
  • 所有任务都是 XL 大小
  • 任务之间没有检查点
  • 没有考虑依赖顺序

验证

开始实现前,确认:

  • 每个任务都有验收标准
  • 每个任务都有验证步骤
  • 任务依赖已识别并正确排序
  • 没有任务触及超过约 5 个文件
  • 主要阶段之间存在检查点
  • 人类已评审并批准计划

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 7 of the 12 instructions most task breakdown skills give in ~2.1k tokens

Counted across 221 of the 230 authors here whose files we hold, read 2026-09-06

  • Write acceptance criteria for every taskhere, and in 32 of 221, across 30 files
  • Add checkpoints every two to three taskshere, and in 18 of 221, across 16 files
  • Schedule high-risk tasks earlyhere, and in 16 of 221, across 14 files
  • Slice work vertically into complete feature pathsin 16 of 221, across 14 files
  • Include verification steps in every taskhere, and in 15 of 221, across 13 files
  • Map dependencies between componentshere, and in 14 of 221, across 12 files
  • Get human approval before implementinghere, and in 14 of 221, across 12 files
  • Enter read-only plan mode before writing codehere, and in 10 of 221
  • Read the spec and codebase before planningin 9 of 221, across 7 files
  • Map the dependency graph before ordering tasksin 8 of 221
  • Break any task touching more than five filesin 8 of 221
  • List each task's dependenciesin 7 of 221

Said here and by no other author read

  • Build one complete feature path per task
  • Record risks and unknowns in the plan

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.