Token导航 LogoToken导航

Agent Skill · 研究检索

breaking-down-stories(breaking down stories 搜索)

breaking-down-stories 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。

breaking-down-stories 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,权限为接触敏感数据;许可证信息还不明确,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 73,543。

总安装 1周安装 12GitHub Stars 5下载 97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
unknown

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:breaking-down-stories(breaking down stories 搜索)
来源仓库:https://github.com/bnadlerjr/dotfiles
仓库路径:skills/breaking-down-stories
安装命令:
npx skills add https://github.com/bnadlerjr/dotfiles --skill breaking-down-stories
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。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/bnadlerjr/dotfiles --skill breaking-down-stories

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。

将用户故事拆解为可在单次配对会话中完成的细粒度任务。支持全工作类别并行思考,产出自包含、可独立执行的任务列表。适用于敏捷开发中的需求细化,每个任务控制在1-4小时内可完成。安装前建议确认项目上下文和交付标准,避免生成脱离实际的孤立任务项。输出结果需结合团队能力评估,确保任务粒度合理且资源匹配可行。

来源仓库内路径:skills/breaking-down-stories

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/bnadlerjr/dotfiles --skill breaking-down-stori…;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

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

breaking-down-stories 怎么安装?

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

breaking-down-stories 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。

breaking-down-stories 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

breaking-down-stories 是免费的吗?

公开资料没有标注明确许可证,收费和商用限制都无法确认。正式使用前请到来源仓库核对授权说明。

breaking-down-stories 用的人多吗?

公开数据是累计安装 1 次,近一周安装 12 次,下载 97 次,GitHub 5 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。

breaking-down-stories 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。
  • 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

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 许可证没有明确标注,能不能商用、能不能再分发都不确定。
  • 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。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要明确授权才能用到公司项目里的场景。
  • 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
breaking-down-stories

breaking down stories 搜索 · 本页

15接触敏感数据未明确通过

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

Breaking Down Stories

Decompose a user story into small, actionable tasks using holistic brainstorming—consider all work categories at once rather than thinking sequentially through the story. Each task should be self-contained enough for a pair to pick up independently.

Quick Start

Given a story, generate a flat task list where each task is small enough to complete in a single focused session:

## Tasks for: [Story Title]

- [ ] Task description
- [ ] Task description -- (context: why this matters)
...

Instructions

  • Each task should be completable in a single pairing session (1-4 hours)
  • Each task should be self-contained—a pair picking it up needs only the task description and context note
  • Add a brief context note when the "why" is not obvious from the task description
  • Include ALL work needed to complete the story (code, tests, builds, design, docs)
  • Keep task descriptions brief—leave room for implementers to work out details
  • Do NOT sequence tasks or assign owners—just brainstorm the work
  • Focus on "what" not "how"

TDD Guardrails

  • NEVER create standalone test tasks—tests are implicit in every implementation task
  • Implementers follow TDD: write failing test first, then implement
  • Valid: "Add User class with validation" (tests written as part of implementation)
  • Invalid: "Add User class" + "Write tests for User class" (redundant split)
  • Invalid: "Set up test fixtures for users" (fixtures created when first test needs them)

Story Quality Guardrails

Before breaking down a story, verify it's ready:

  • Missing acceptance criteria: Ask for clarification or create a "Define acceptance criteria for [feature]" task
  • Too vague: If you can't identify concrete work, the story needs refinement—flag this to the user
  • Too large (spans multiple epics): Suggest splitting into smaller stories first
  • Feature-sized input (spans multiple layers, multiple user roles, or lacks acceptance criteria): Suggest invoking slicing-elephant-carpaccio to produce thin vertical slices first, then writing-agile-stories for each slice, then return here for task decomposition
  • Missing user/value: If "who" or "why" is unclear, ask before proceeding

Input Classification

Before decomposing, classify the input:

SignalClassificationAction
Has acceptance criteria, single user role, bounded scopeStoryProceed with task decomposition
Spans multiple layers, multiple roles, no acceptance criteriaFeatureSuggest slicing-elephant-carpaccio first
References an epic or multiple storiesEpicSuggest splitting into stories with writing-agile-stories first
Single technical concern, already smallTaskNo decomposition needed—inform the user

If uncertain, ask the user whether to slice first or proceed directly.

For Complex Stories

  • If the story spans multiple systems or has unclear boundaries, invoke the thinking-patterns skill with atomic to decompose requirements first
  • If verification keeps finding gaps, invoke the thinking-patterns skill with self-consistency to evaluate completeness from multiple perspectives

For Modification-Heavy Stories

  • If the story primarily modifies existing code (vs. greenfield), quickly scan relevant modules to identify integration points
  • Keep reconnaissance lightweight—don't get pulled into implementation details

Handling Open Questions

  • If the story has open questions, create an explicit task: "Resolve: [question]"
  • Skip tasks that depend on open questions—don't speculate on work that may not exist
  • Note dependencies in the output so reviewers know the list is intentionally incomplete

Workflow

  1. Gather the Story

- If input is a ticket reference (e.g., "PROJ-123"), use available tools to fetch details (e.g., jira issue view PROJ-123) - If input is prose, use it directly - Identify: who is the user, what do they want, why do they want it

  1. Brainstorm Tasks Holistically

- Consider all work categories at once: code changes, configuration, documentation, deployment, infrastructure - Include non-obvious work: "Update build script", "Add migration", "Update API docs" - Don't think sequentially through the story—capture all work regardless of order

  1. Right-Size Each Task

- Split tasks that span multiple concerns or would take multiple sessions - Combine trivial tasks that are always done together - Each task should be independently completable - Target 1-4 hours per task (one pairing session) - Add a context note when the task's purpose is not self-evident

  1. Verify Completeness

- Walk through the story end-to-end - Ask: "If all these tasks are done, is the story shippable?" - Add any missing tasks

  1. Check Task Quality

- Scan for any test-related tasks → remove them (tests are implicit) - Scan for "and" in task names → likely compound tasks, split them - Scan for vague tasks like "Implement feature" → add specificity - Scan for tasks missing context → add brief context notes where the "why" is not obvious - Confirm task count is reasonable (5-15 typical; fewer suggests tasks too large, more suggests story too large) - If task count exceeds 15, the input may be feature-sized → revisit Input Classification

Anti-Patterns

Avoid these common mistakes:

Bad TaskProblemBetter Task
"Write tests for login"Separated test taskRemove—tests are implicit
"Set up test factories"Test infrastructure taskRemove—created when first needed
"Implement authentication"Too vague"Add JWT token generation on login"
"Update code and fix bug"Compound taskSplit into two tasks
"Use bcrypt for passwords"Implementation detail"Add secure password hashing"
"Research options"Not actionable"Resolve: which auth library to use"
"Add password reset endpoint"Missing context for non-obvious purpose"Add password reset endpoint -- (entry point for the reset flow)"

Example

Input Story:

## Story: User Resets Forgotten Password

When users cannot remember their password or suspect their account may be
compromised, they need a secure way to reset their credentials and regain
access. The process must verify their identity while remaining accessible.

### Context
Available from the login page when a user cannot authenticate.

### Acceptance Criteria

#### Scenario: Successful password reset
- Given a user with a verified email address
- When they request a password reset
- Then they receive an email with a secure reset link
- And the link expires after 1 hour
- And they can set a new password using the link

#### Scenario: Reset for unregistered email
- Given the email address is not in the system
- When someone requests a reset for that email
- Then the same "check your email" message is shown
- And no email is sent (prevents account enumeration)

#### Scenario: Expired reset link
- Given a user received a reset link more than 1 hour ago
- When they try to use the link
- Then they are informed the link has expired
- And they can request a new reset link

#### Scenario: Multiple reset requests
- Given a user has requested reset 3 times in 10 minutes
- When they request another reset
- Then they are asked to wait before trying again

Output:

## Tasks for: User Resets Forgotten Password

- [ ] Add password reset request endpoint with email lookup
- [ ] Create reset token generation with secure random and 1-hour expiration
- [ ] Store reset tokens with user association and expiration timestamp
- [ ] Add email template for password reset link
- [ ] Integrate email sending for reset requests
- [ ] Add password reset confirmation endpoint with token validation
- [ ] Invalidate token after successful password change -- (prevents token reuse)
- [ ] Add rate limiting: max 3 requests per 10 minutes per email -- (prevents brute force)
- [ ] Return consistent response regardless of email existence -- (prevents account enumeration)
- [ ] Add error handling for expired tokens with re-request guidance
- [ ] Update login page with "Forgot password?" link
- [ ] Add password validation on reset -- (same rules as registration)

Output Format

Output tasks as a simple list:

## Tasks for: [Story Title]

- [ ] Task description
- [ ] Task description -- (context: why this matters)
- [ ] Task description
...

Keep descriptions to one line. No estimates, no assignments, no implementation details. Add context notes sparingly—only when the "why" is not obvious from the task description.

Success Criteria

A well-decomposed story produces tasks where:

  • Every task is independently completable in a single pairing session (1-4 hours)
  • No task is a standalone test task (TDD is implicit)
  • The task list covers ALL work needed to ship the story
  • A pair can pick up any task using only its description and context note
  • Task count is 5-15 (fewer = tasks too large, more = story too large or feature-sized)
  • Walking through all completed tasks would make the story shippable

Related Skills

Upstream (use before this skill)

  • slicing-elephant-carpaccio — Slice a feature into thin vertical increments. Use when input is feature-sized (see Input Classification). Each slice becomes a story candidate.
  • writing-agile-stories — Write BDD stories with acceptance criteria. Use when input lacks acceptance criteria or clear scope.

Downstream (use after this skill)

  • implementation-planning — Create detailed implementation plans for individual tasks from the task list.

Typical workflow

Feature → slicing-elephant-carpaccio → writing-agile-stories → breaking-down-stories → implementation-planning

平台分布

Codex

36.52%
按下载量换算35

Claude

27.65%
按下载量换算27

Cursor

18.44%
按下载量换算18

Gemini CLI

8.3%
按下载量换算8

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills