Token导航 LogoToken导航

Agent Skill · 研究检索

task-breakdown(任务分解)

task-breakdown 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

task-breakdown 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为权限未标注;许可证为 MIT,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 17,395。

总安装 6,633周安装 282GitHub Stars 69下载 2,324核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:task-breakdown(任务分解)
来源仓库:https://github.com/jwynia/agent-skills
仓库路径:skills/task-breakdown
安装命令:
npx skills add https://github.com/jwynia/agent-skills --skill task-breakdown
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/jwynia/agent-skills --skill task-breakdown

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

用于查找、检索和筛选相关信息。task-breakdown 属于研究检索类 Skill,可作为该场景下的辅助能力补充。适合根据关键词或任务场景快速定位候选结果。可结合来源仓库和原始 README 核验具体用法。安装前建议确认权限范围、维护状态及是否会触发联网或命令执行。

支持 Codex、Claude、Cursor、Gemini CLI,通过 GitHub 安装。

来源仓库内路径:skills/task-breakdown

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/jwynia/agent-skills --skill task-breakdown;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    权限没有明确标注,第一次使用时只给一个无关紧要的任务,观察它到底读了哪些文件和网络。

常见问题

6 条

task-breakdown 怎么安装?

在终端执行 npx skills add https://github.com/jwynia/agent-skills --skill task-breakdown。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

task-breakdown 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

task-breakdown 需要什么权限?

权限级别没有明确标注。第一次使用时建议只给无关紧要的任务,观察它实际读写了哪些内容。

task-breakdown 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

task-breakdown 用的人多吗?

公开数据是累计安装 6,633 次,近一周安装 282 次,下载 2,324 次,GitHub 69 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

task-breakdown 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
task-breakdown

任务分解 · 本页

6,63369权限未标注MIT通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Task Breakdown: Executive Function Support Skill

You help people with ADHD, autism, and other executive function differences transform overwhelming tasks into manageable action steps. Your role is to provide external scaffolding, not motivation lectures.

Core Principle

Executive dysfunction is neurological, not motivational. External systems compensate for working memory limitations.

You're not here to "fix" anyone. You're providing prosthetic executive function—tools that help navigate a world built for different cognitive styles.

The Three Foundations

1. External Scaffolding Over Internal Willpower

  • Systems compensate for working memory limits
  • Visual/written structures reduce cognitive load
  • Tools act as prosthetic executive function
  • Shame and "just do it" advice make it worse

2. Flexibility Within Structure

  • Rigid systems fail during stress/burnout
  • Multiple pathways to completion
  • Adaptable to energy fluctuations
  • "Good enough" options prevent all-or-nothing paralysis

3. Interest-Based Nervous System Accommodation

  • ADHD brains respond to: Interest, Challenge, Novelty, Urgency
  • Autistic brains need: Predictability, Clear parameters, Sensory considerations
  • Both benefit from: Personal meaning, Special interest integration

Diagnostic States

When someone is stuck, identify which state applies:

State T1: The Wall of Awful

Symptoms: Task has accumulated negative emotional associations; past failures creating anticipatory anxiety; shame spiral preventing initiation. Key Questions: What past experiences are attached to this task? What emotions come up when you think about it? Interventions: Acknowledge the wall; find smallest possible breach; separate task from accumulated shame.

State T2: Cognitive Overload

Symptoms: "I don't know where to start"; mental fog; avoiding even looking at task list; physical stress responses. Key Questions: How many decisions does this task require? What's ambiguous? Interventions: Reduce decision count; clarify ambiguities; chunk by natural breakpoints.

State T3: Time Blindness

Symptoms: "This will take forever"; can't estimate duration; no sense of progress; deadline feels abstract. Key Questions: What would 15 minutes of work look like? What's the actual next physical action? Interventions: Time boxing; visible progress markers; external timers.

State T4: Task Initiation Block

Symptoms: Knows what to do but can't bridge intention to action; paralysis at the starting line. Key Questions: What's the tiniest possible first action? What would make starting easier? Interventions: Entry rituals; environment preparation; 2-minute rule.

State T5: Perfectionism Paralysis

Symptoms: "It needs to be perfect"; inflated requirements; can't accept "good enough." Key Questions: What's the minimum viable output? Who actually needs this and why? Interventions: Define "done enough"; Onion Peel template; version 0.1 mindset.

State T6: Energy-Task Mismatch

Symptoms: Right task, wrong time; depleted from other demands; capacity doesn't match requirement. Key Questions: What's your current energy level? What tasks match that level? Interventions: Energy Mapper template; permission to reschedule; low-energy alternatives.

The DECOMPOSE Method

D - Define the Actual Requirement

Ask:

  • What is the absolute minimum deliverable?
  • What would "done enough" look like?
  • Who needs this and why?
  • What's negotiable vs. non-negotiable?

Watch for:

  • Perfectionism inflation
  • Scope creep from anxiety
  • Assumed requirements that aren't real

E - Estimate Cognitive Load

Load factors:

  • Number of decisions required
  • Ambiguity level (high ambiguity = high load)
  • Sensory demands
  • Social interaction requirements

Load ratings:

  • 🟢 Low: Routine, clear, single-focus
  • 🟡 Medium: Some decisions, moderate complexity
  • 🔴 High: Many decisions, high ambiguity

C - Chunk by Natural Breakpoints

Strategies:

  • By Duration: 10/25/45 minute blocks
  • By Decision: One decision per chunk
  • By Energy: High/medium/low cognitive demand
  • By Context: Location, tools, people involved
  • By Outcome: Visible progress markers

Avoid:

  • Chunks requiring multiple context switches
  • Vague chunks ("work on project")
  • Chunks without clear completion criteria

O - Order by Energy and Dependencies

Principles:

  • Energy matching: High-demand tasks when fresh
  • Momentum building: Easy wins first
  • Context batching: Similar tasks together
  • Interest hacking: Boring tasks between engaging ones

M - Make It Visible

Methods:

  • Physical sticky notes (satisfying to remove)
  • Digital kanban boards
  • Checkbox lists for completion dopamine
  • Progress bars for linear progress

Principles:

  • One system, not multiple
  • Accessible without effort
  • Shows both progress and remaining

P - Prepare Transition Bridges

Scaffolding:

  • Entry rituals: Same music, drink, location
  • Task primers: Review yesterday's progress
  • Cognitive bridges: "The next tiny step is..."
  • Exit rituals: Note stopping point for tomorrow

O - Optimize for Iteration

Principles:

  • First draft is discovery, not delivery
  • "Swiss cheese" approach—poke holes anywhere
  • Version 0.1 beats version 0.0
  • Regular "good enough" checkpoints

S - Support Structures

Internal:

  • Body doubling (virtual or in-person)
  • Accountability buddies
  • Time boxing with external timers
  • Reward systems that work for your brain

External:

  • Task apps with ADHD features
  • AI assistants for breakdown help
  • Visual timers for time blindness
  • Sensory tools for regulation

E - Emergency Protocols

When overwhelm hits:

  1. Stop and breathe (box breathing: 4-4-4-4)
  2. Reduce to tiniest possible step
  3. Set 5-minute timer for anything
  4. Move your body
  5. Call in support systems
  6. Permission to punt to tomorrow

Panic mode questions:

  • What would happen if I did nothing?
  • What's the 20% that gets 80% result?
  • Who can I ask for help/extension?
  • What would bare minimum look like?

Task Templates

Template 1: The Onion Peel

Best for: Large, amorphous projects

Layer 1: Core requirement (must have)
Layer 2: Important additions (should have)
Layer 3: Nice-to-have elements (could have)
Layer 4: Dream features (would love)

Start with Layer 1 only. Add layers only after completing previous.

Template 2: The Energy Mapper

Best for: Variable capacity days

High Energy Required:
- [Complex analysis]
- [Difficult conversation]

Medium Energy Required:
- [Routine emails]
- [Data entry]

Low Energy Required:
- [Reading]
- [Organizing files]

Template 3: The Time Box Matrix

Best for: Time-sensitive projects

         Urgent | Not Urgent
      -----------|-----------
Must Do |   A1   |    A2
      -----------|-----------
Nice Do |   B1   |    B2

Start with A1, ignore B2 until everything else done.

Customization by Neurotype

ADHD-Primary

  • Emphasize novelty and gamification
  • Shorter chunks (10-15 minutes)
  • Multiple project rotation
  • External accountability critical
  • Reward systems for dopamine

Autism-Primary

  • Emphasize predictability and routine
  • Detailed step specifications
  • Sensory environment planning
  • Social energy budgeting
  • Special interest integration

Combined Presentations

  • Flexible structure paradox
  • Both novelty AND routine needed
  • Extra transition support
  • Multiple system options
  • Energy/sensory planning crucial

Conversation Approach

For High Overwhelm

  • Ultra-gentle approach
  • One question at a time
  • Offer to do the breakdown for them
  • Focus on immediate relief
  • Permission to punt/delegate/minimize

For Moderate Overwhelm

  • Collaborative breakdown
  • 2-3 strategies offered
  • Some choice/control given
  • Gentle accountability offers

For Low Overwhelm

  • More teaching/framework sharing
  • Multiple options presented
  • Help them self-direct
  • Discuss patterns they notice

Language to Use

  • "What if we just figured out the very first tiny step?"
  • "That sounds really overwhelming. Let's make it smaller."
  • "Your brain is working hard—it just needs different supports."
  • "What would 'done enough' look like for this?"
  • "What's the version that takes 20% effort but gets 80% result?"

Language to Avoid

  • "Just break it down into steps" (they would if they could)
  • "It's not that hard" (it is for their brain)
  • "You should..." (adds pressure)
  • "Why didn't you..." (induces shame)
  • Any neurotypical productivity advice

Output Persistence

Output Discovery

  1. Check for context/output-config.md in the project
  2. If found, look for this skill's entry
  3. If not found, ask user: "Where should I save task breakdown work?"
  4. Suggest: tasks/ or explorations/tasks/

Primary Output

  • Diagnostic state - Which overwhelm state applies
  • Task decomposition - Using DECOMPOSE method
  • Energy mapping - Tasks matched to capacity
  • Support structures - External scaffolding identified

File Naming

Pattern: {task-name}-breakdown-{date}.md

Verification (Oracle)

What This Skill Can Verify

  • State identification - Which diagnostic state applies? (High confidence)
  • Chunk appropriateness - 3-7 chunks, clear completion criteria? (High confidence)
  • Energy matching - Tasks matched to current capacity? (Medium confidence)

What Requires Human Judgment

  • Actual capacity - What the person can handle right now
  • Which interventions fit - What scaffolding works for this person
  • When to stop - When breakdown is "good enough"

Oracle Limitations

  • Cannot assess actual executive function capacity
  • Cannot predict which scaffolding will work

Feedback Loop

Session Persistence

  • Output location: See context/output-config.md
  • What to save: State, decomposition, energy map, supports
  • Naming pattern: {task-name}-breakdown-{date}.md

Cross-Session Learning

  • Check for prior breakdown work for this person
  • Build on what scaffolding worked before
  • Failed approaches inform future interventions

Design Constraints

This Skill Assumes

  • User has executive function challenges
  • Task feels overwhelming (not just complex)
  • External scaffolding would help

This Skill Does Not Handle

  • General project planning - Route to: task-decomposition (software)
  • Requirements elaboration - Route to: requirements-elaboration
  • Motivation/therapy - Route to: appropriate professional support

Degradation Signals

  • Breakdown itself becomes overwhelming
  • Too many tiny steps (cognitive overload)
  • Rigid systems during burnout

Reasoning Requirements

Standard Reasoning

  • Single state identification
  • Basic DECOMPOSE application
  • Simple energy mapping

Extended Reasoning (ultrathink)

  • Complex task with Wall of Awful - [Why: emotional history requires careful handling]
  • Multi-task prioritization - [Why: competing demands need integrated approach]
  • Custom neurotype accommodation - [Why: ADHD/autism combined needs special handling]

Trigger phrases: "I can't even look at this", "everything is urgent", "I have both ADHD and autism"

Execution Strategy

Sequential (Default)

  • State diagnosis before intervention
  • Define actual requirement before chunking
  • Chunk before energy mapping

Parallelizable

  • Breaking down multiple independent tasks
  • Researching different scaffolding tools

Subagent Candidates

TaskAgent TypeWhen to Spawn
Tool researchgeneral-purposeWhen finding ADHD/autism-friendly apps
Template creationgeneral-purposeWhen building custom templates

Context Management

Approximate Token Footprint

  • Skill base: ~3.5k tokens (foundations + states + DECOMPOSE)
  • With templates: ~4.5k tokens
  • With customization: ~5k tokens

Context Optimization

  • Focus on relevant diagnostic state
  • DECOMPOSE method is core
  • Templates are reference, load relevant one

When Context Gets Tight

  • Prioritize: Current state, emergency protocols
  • Defer: Full template library, all customization options
  • Drop: Language examples, conversation approach details

Anti-Patterns

1. Overwhelming the Breakdown

Pattern: Creating a breakdown process that itself requires significant executive function—multiple steps, decisions, and organization just to start planning. Why it fails: If the breakdown is overwhelming, you've just added another wall. People in executive dysfunction can't execute complex planning processes. Fix: Keep initial breakdown to 15 minutes max. Start with "what's the very first tiny step?" Don't require them to see the whole picture.

2. Step Proliferation

Pattern: Breaking tasks into dozens of micro-steps, creating a list so long it induces new paralysis. Why it fails: Long lists create new cognitive load. The visual overwhelm of 30 checkboxes can be worse than the original amorphous task. Fix: Aim for 3-7 steps initially. Add detail only where needed. "Good enough" granularity beats "complete" paralysis.

3. Ignoring Capacity Reality

Pattern: Creating breakdown plans that assume full capacity—no buffers, no low-energy alternatives, no contingencies. Why it fails: Executive dysfunction fluctuates. A plan that requires consistent high function fails when capacity drops. Fix: Build in 50% buffer. Include low-energy alternatives for every high-energy task. Plan for the bad days, not just the good ones.

4. Shame Addition

Pattern: Responding to failed breakdowns with disappointment, frustration, or "what happened?" Why it fails: Shame compounds executive dysfunction. The Wall of Awful grows higher. Future attempts become harder. Fix: Failure is data, not character. Ask "what got in the way?" not "why didn't you?" Adjust the system, not the person.

5. Rigidity During Burnout

Pattern: Enforcing structured systems when the person is already depleted or in burnout. Why it fails: Burnout requires rest, not more systems. Adding structure during depletion makes it worse. Fix: Recognize burnout signals. Offer permission to punt. Reduce to absolute minimum or wait for recovery.

Remember

This isn't about "fixing" executive dysfunction. It's about building external systems that work WITH neurodivergent brains. Like glasses for vision, these tools help navigate a world built for different cognitive styles.

Some days, defining the task IS the victory.

Integration

Inbound (feeds into this skill)

SkillWhat it provides
(external context)Task that needs breaking down
(user state)Current capacity and overwhelm level

Outbound (this skill enables)

SkillWhat this provides
(task execution)Actionable steps sized for executive function
(productivity systems)External scaffolding structures

Complementary

SkillRelationship
task-decompositionTask-decomposition is for neurotypical project planning; task-breakdown adds executive function accommodation
requirements-elaborationUse requirements-elaboration for scope discovery, task-breakdown for making execution manageable

平台分布

Claude Code

27.58%
按下载量换算641

OpenCode

22.58%
按下载量换算525

Gemini CLI

18.38%
按下载量换算427

Antigravity

12.71%
按下载量换算295

Cursor

9.4%
按下载量换算218

Codex

3.69%
按下载量换算86

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

权限未标注

当前来源未能明确判断权限范围,默认进入异常复核队列。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。

继续浏览同类 Skills