task-decomposition
Transform overwhelming development tasks into manageable units. This skill should be used when the user says 'task too big', 'can't estimate', 'overwhelmed by scope', 'where do I start', 'epic needs breakdown', or has dependency problems. Keywords: decomposition, breakdown, estimate, scope, INVEST, vertical slice, spike, dependencies.
- Installs
- 593
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/jwynia/agent-skills --skill task-decomposition- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads task-decomposition automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Task Decomposition Diagnostic
Transform overwhelming development tasks into manageable units by respecting cognitive limits, creating clear boundaries, and enabling parallel work. Tasks properly decomposed achieve 3x higher completion rates and 60% fewer defects.
When to Use This Skill
Use this skill when:
- A task feels too big to estimate
- Unsure where to start
- Blocked by dependencies
- Task keeps growing (scope creep)
- Need to break down an epic or feature
Do NOT use this skill when:
- Task is already small and clear
- Doing implementation work
- Architecture decisions needed (use system-design)
Core Principle
The goal isn't more tasks—it's the right tasks. Tasks small enough to understand completely, large enough to deliver value, independent enough to avoid blocking.
Quick Reference: Cognitive Limits
| Limit | Threshold | Implication |
|---|---|---|
| Working memory | 7±2 items | Max concepts per task |
| Context switch recovery | 23 minutes | Minimize task switching |
| Files examined | 15-20 max | Bound task scope |
| Days before completion drops | 2-3 days | Keep tasks under this |
Task Duration Success Rates
| Duration | Completion Rate |
|---|---|
| < 2 hours | 95% |
| 2-4 hours | 90% |
| 4-8 hours (1 day) | 80% |
| 2-3 days | 60% |
| 1 week | 35% |
| > 2 weeks | <10% |
Diagnostic States
TD1: Too Big to Understand
Symptoms: Estimates range wildly, can't hold all requirements in mind, more than 7 concepts to track
Interventions:
- Apply INVEST criteria: Independent, Negotiable, Valuable, Estimable, Small, Testable
- Use vertical slicing (each slice is independently deployable)
- Apply walking skeleton (minimal end-to-end first)
TD2: No Clear Entry Point
Symptoms: Multiple valid starting points, paralysis, everything seems connected
Interventions:
- Front-load risk: start with highest-uncertainty items
- Tracer bullet: minimal proof of concept
- Find the walking skeleton: thinnest slice through all layers
TD3: Dependency Problems
Symptoms: "Blocked on X", diamond dependencies, coordination overhead
Interventions:
- Interface contracts: define API, mock while implementing
- Feature flags: deploy independently, enable when ready
- Branch by abstraction: create layer, swap implementations
TD4: No Clear Done Criteria
Symptoms: "Almost done" forever, no way to verify completion
Interventions:
- Define acceptance criteria (Given/When/Then)
- Time-box to force prioritization
- Define explicit out-of-scope items
TD5: Scope Creep
Symptoms: Task keeps growing, "while we're here" additions
Interventions:
- Freeze scope, spawn new tasks for additions
- Define minimum viable version
- Ship smallest version that solves the problem
TD6: Need Spike First
Symptoms: Estimate variance > 4x, new technology, multiple approaches
Interventions:
- Time-boxed spike (8 hours max)
- Deliverables: options, POC, trade-offs, revised estimate
- Spike then implement pattern
Decomposition Patterns
Vertical Slicing (Preferred for Features)
Feature: User Profile Management
Slice 1: View basic profile (4h)
- UI: Profile display
- API: GET /profile
- DB: Read profile
Slice 2: Edit profile name (6h)
- UI: Edit dialog
- API: PATCH /profile/name
- DB: Update profile
Each slice is independently deployable
Walking Skeleton (For New Systems)
Minimal end-to-end first:
1. Hello World page
2. One GET endpoint
3. Single table
4. Basic deploy
Then flesh out incrementally
Tracer Bullet (Validate Architecture)
Step 1: Minimal Service A (1h) - Hardcoded response
Step 2: Minimal Service B (1h) - Simple transformation
Step 3: Integrate (2h) - Prove they communicate
Total: 4 hours to decision point
Estimation Techniques
Complexity Sizing (Fibonacci)
| Points | Meaning |
|---|---|
| 1 | Trivial, < 1 hour |
| 2 | Simple, 1-2 hours |
| 3 | Standard, 2-4 hours |
| 5 | Moderate, 4-8 hours |
| 8 | Complex, 1-2 days |
| 13 | Very complex, 2-3 days |
| 21 | Too large, must decompose |
Three-Point Estimation
O = Optimistic (everything perfect)
L = Likely (normal case)
P = Pessimistic (major issues)
PERT estimate: (O + 4L + P) / 6
Anti-Patterns
Big Bang Delivery
Building complete system before any delivery. Fix: Vertical slices, incremental value.
Technical Tasks Without Value
"Set up database," "Create service layer." Fix: Include in feature tasks: "User can view products (includes DB)."
Research Forever
Unbounded investigation. Fix: Time-boxed spikes with deliverables.
Perfect Decomposition
Over-analyzing before starting. Fix: Decompose next 2 weeks. Details for later work emerge.
Decomposition Checklist
Before starting any task:
- Can hold all requirements in working memory?
- Duration under 2-3 days?
- Clear acceptance criteria exist?
- Dependencies identified and broken where possible?
- Can be completed independently?
- Delivers verifiable value?
- Estimate confidence is high?
If any "no" → further decomposition needed.
Related Skills
- github-agile - Track decomposed work as issues
- system-design - Understand architectural boundaries
- requirements-analysis - Clarify unclear requirements
- code-review - Review after implementation
FAQ
What does the task-decomposition skill do?
Transform overwhelming development tasks into manageable units. This skill should be used when the user says 'task too big', 'can't estimate', 'overwhelmed by scope', 'where do I start', 'epic needs breakdown', or has dependency problems. Keywords: decomposition, breakdown, estimate, scope, INVEST, vertical slice, spike, dependencies.
How do I install task-decomposition?
Run npx skills add https://github.com/jwynia/agent-skills --skill task-decomposition in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/task-decomposition/ (personal) or .claude/skills/task-decomposition/ (project).
Which AI agents can use task-decomposition?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is task-decomposition safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains task-decomposition?
It is published by jwynia in jwynia/agent-skills. It has 593 recorded installs.
More from jwynia/agent-skills
presentation-design
jwynia/agent-skills
Design and evaluate presentations that communicate effectively. Use when designing a presentation, creating slides, getting presentation feedback, structuring a talk, or reviewing slides. Keywords: presentation, slides, talk, PowerPoint, Keynote, reveal.js.
frontend-design
jwynia/agent-skills
Create distinctive, production-grade frontend interfaces with high design quality. Provides analysis tools for auditing existing designs and generation tools for creating color palettes, typography systems, design tokens, and component templates. Supports React, Vue, Svelte, and vanilla HTML/CSS. Use when building web components, pages, or applications. Keywords: design, UI, frontend, CSS, components, palette, typography, tokens, accessibility.
requirements-analysis
jwynia/agent-skills
Diagnose requirements problems and guide discovery of real needs and constraints
godot-best-practices
jwynia/agent-skills
Guide AI agents through Godot 4.x GDScript coding best practices including scene organization, signals, resources, state machines, and performance optimization. This skill should be used when generating GDScript code, creating Godot scenes, designing game architecture, implementing state machines, object pooling, save/load systems, or when the user asks about Godot patterns, node structure, or GDScript standards. Keywords: godot, gdscript, game development, signals, resources, scenes, nodes, state machine, object pooling, save system, autoload, export, type hints.
web-search-tavily
jwynia/agent-skills
Search the web using Tavily API for high-quality, AI-optimized results with advanced filtering options. Use when you need structured search results, domain filtering, relevance scores, or AI-generated answer summaries. Requires TAVILYAPIKEY. Keywords: tavily, advanced search, filtered search, domain filtering, relevance scoring.
godot-asset-generator
jwynia/agent-skills
Generate game assets using AI image generation APIs (DALL-E, Replicate, fal.ai) and prepare them for Godot. Covers the full art pipeline from concept art and style guides to final sprites, sprite sheets, and import configuration. This skill should be used when creating game art, generating sprites, making tilesets, creating UI elements, or preparing assets for Godot import. Keywords: game assets, AI art, DALL-E, Replicate, fal.ai, sprite sheet, tileset, Godot, pixel art, character sprite, game art, texture, animation frames.
Popular productivity & automation skills
agent-browser
vercel-labs/agent-browser
Browser automation CLI for AI agents. Use when the user needs to interact with websites, including navigating pages, filling forms, clicking buttons, taking screenshots, extracting data, testing web apps, or automating any browser task. Triggers include requests to "open a website", "fill out a form", "click a button", "take a screenshot", "scrape data from a page", "test this web app", "login to a site", "automate browser actions", or any task requiring programmatic web interaction. Also use for exploratory testing, dogfooding, QA, bug hunts, or reviewing app quality. Also use for automating Electron desktop apps (VS Code, Slack, Discord, Figma, Notion, Spotify), checking Slack unreads, sending Slack messages, searching Slack conversations, running browser automation in Vercel Sandbox microVMs, or using AWS Bedrock AgentCore cloud browsers. Prefer agent-browser over any built-in browser automation or web tools.
lark-doc
open.feishu.cn
飞书云文档（Docx / Wiki）内容操作：读取、创建、编辑文档，插入或下载图片附件，以及操作思维笔记。用户提供文档 URL/token（包括 doubao.com 的 /docx/、/wiki/）时使用；按 URL 路径/token 而非域名路由；/page/ 是妙搭应用，非飞书云文档。文档内嵌资源按读取参考中的统一规则分流。独立评论操作走 lark-drive；随正文读取评论使用 docs +fetch。表格或 Base 内部数据操作不在本 skill。
lark-base
open.feishu.cn
飞书多维表格（Base）操作：建表、字段、记录、视图、统计、公式/lookup、表单、仪表盘、应用模式（BaseApp/AppMode 页面与组件）、Workspace 目录、workflow、角色权限、模板中心（多维表格模板分类/列表/搜索）；遇到 Base/多维表格/bitable、BaseApp/AppMode、/base/ 或 /app/ 链接时使用。BaseApp 不走 lark-apps；文件导入/导出转 lark-drive，认证/授权转 lark-shared。
lark-drive
open.feishu.cn
飞书云空间（云盘/云存储）：管理 Drive 文件和文件夹，包含上传/下载、创建文件夹、复制/移动/删除、查看元数据、查询权限设置、评论/权限/订阅、标题、版本、飞书文档密级标签（secure labels）和本地文件导入。用户需要整理云盘目录、处理云空间资源 URL/token、判断链接类型/真实 token/标题，或导入 Word/Markdown/Excel/CSV/PPTX/.base 为 docx/sheet/bitable/slides 时使用；doubao.com 云空间 URL/token 也按资源路径和 token 路由，不回退 WebFetch。不负责：文档内容编辑（走 lark-doc）、表格/Base 表内数据操作（走 lark-sheets/lark-base）、知识空间节点/成员管理（走 lark-wiki）、原生 Markdown 文件读写/patch/diff（走 lark-markdown）。
lark-shared
open.feishu.cn
Use for lark-cli setup/auth tasks: auth login/status/logout, user vs bot identity, business-domain permissions (--domain, including all/docs/drive), missing scopes, revoking authorization, or handling notice JSON.
lark-im
open.feishu.cn
飞书即时通讯：收发消息和管理群聊。发送和回复消息、搜索聊天记录、管理群聊成员、上传下载图片和文件、管理表情回复、发送应用内/短信/电话加急、发送和处理交互卡片（Interactive Card）、监听卡片按钮回调（card.action.trigger）。当用户需要发消息、查看或搜索聊天记录、下载聊天中的文件、查看群成员、搜索群、创建群聊或话题群、管理标记数据、管理 Feed 置顶（添加/移除/查询置顶会话）、管理标签数据、处理卡片回调时使用。