Token导航 LogoToken导航

Agent Skill · 研究检索

cdd-plan(CDD 计划)

cdd-plan 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。

cdd-plan 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,权限为需要联网;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 69,971。

总安装 214周安装 9GitHub Stars 1下载 75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:cdd-plan(CDD 计划)
来源仓库:https://github.com/ruphware/cdd-skills
仓库路径:skills/cdd-plan
安装命令:
npx skills add https://github.com/ruphware/cdd-skills --skill cdd-plan
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/ruphware/cdd-skills --skill cdd-plan

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。

cdd-plan 用于将目标需求转化为可执行的 TODO 步骤,保留现有模板同时补充缺失部分,适合在 Codex、Claude、Cursor、Gemini CLI 中制定实施路线图时使用。它读取 AGENTS.md、README、specs 等文件作为 truth source,确保计划贴合实际上下文。

非平凡工作推荐包含 Goal、Constraints、Tasks、Implementation notes 与 Automated checks 等节。使用前应确认 spec 文档完整,避免因信息不全导致步骤不可测。cdd-plan 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/cdd-plan

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/ruphware/cdd-skills --skill cdd-plan;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

    它需要需要联网,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

cdd-plan 怎么安装?

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

cdd-plan 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。

cdd-plan 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

cdd-plan 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。

cdd-plan 用的人多吗?

公开数据是累计安装 214 次,近一周安装 9 次,下载 75 次,GitHub 1 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。

cdd-plan 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

  • 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

  • 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
cdd-plan

CDD 计划 · 本页

2141需要联网未明确通过

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

CDD Plan (explicit-only)

Treat the target repo’s CDD contract files as the source of truth:

  • AGENTS.md
  • README.md
  • TODO.md (and/or TODO-*.md)
  • docs/specs/prd.md
  • docs/specs/blueprint.md
  • docs/INDEX.md (if present)

Runnable TODO step contract

For any new or rewritten execution step, produce an implementation-ready step rather than a placeholder summary.

  • Preserve the repo's existing Step template when possible, but add missing sections when the current template would leave the step underspecified.
  • Preferred section set for non-trivial work:

- Goal - Constraints - Tasks - Implementation notes - Automated checks - UAT

  • A runnable step is decision-complete: the implementer can execute it without reopening PRD/Blueprint to discover missing product, architecture, sequencing, or validation decisions.
  • Each Tasks bullet must name:

- the target boundary or subsystem - the exact change to make - the output artifact, contract, or behavior that must result - any must-preserve invariant or evidence requirement when relevant

  • Use Implementation notes for file/symbol hints, interface or schema changes, ordering constraints, migration notes, snapshot/audit requirements, and other coding-critical detail that would otherwise be lost in a short task list.
  • Do not leave essential implementation detail only in the surrounding chat. Put it in the TODO step.
  • Split work into separate steps when it crosses distinct hard gates, migration boundaries, rollback surfaces, or independently testable subsystems.

Interactive planning contract

Planning in this skill is interactive, review-driven, and continuously refined.

  • Start in planning mode when the runtime supports a native read-only or plan mode. If it does not, emulate that behavior by staying read-only until the user approves applying the plan.
  • Review the codebase before and during planning. Audit the relevant code, tests, entrypoints, configs, and current TODO surfaces so the plan is grounded in the actual implementation, not just the docs or user prompt.
  • Treat clarification as a way to resolve the right assumptions, goals, and implementation paths. Do not ask preference questions that do not materially affect the plan.
  • Ask at most one substantive clarification or decision question per message.
  • Keep refining the execution plan as new evidence appears. After each user answer or new repo finding, update boundaries, sequencing, assumptions, and validation requirements before continuing.
  • Keep messages easy to scan: concise, no fluff, and use lightweight Markdown emphasis such as **bold** and *italics* when helpful. Do not depend on color.
  • For every clarification or decision message, put the choices at the bottom under a final **Options** section:

- offer 2-4 concrete options grounded in the repo context - put the recommended option first and mark it clearly - prefix every option label with a visible selector in the label itself so plan-mode UIs still show a selectable key - default to letters: A., B., C. - use numbers only when the surrounding context is already numeric and that would be clearer - keep each option short and action-oriented - avoid open-ended options unless a free-form value is truly required - when practical, tell the user they can reply with just the selector

Flow (approval-gated)

  1. Read the contract files above, any linked sub-specs, and the relevant codebase surfaces.

- Review the current implementation before planning: affected modules, entrypoints, tests, manifests, configs, and existing TODO steps. - Identify the likely boundaries, risks, and validation surfaces for the requested change.

  1. Ask for the change request only if it is not already clear from the user prompt.
  2. Before drafting edits, identify only the blocking or plan-shaping clarifications that would materially change assumptions, goals, implementation paths, file placement, sequencing, or approval boundaries.
  3. Ask clarifying questions one at a time using the interaction contract above.
  4. If any material assumption would remain after the answers, list only those material assumptions and ask the user to confirm or correct them before continuing.
  5. If only minor defaults remain, disclose them briefly in the plan and proceed without blocking.
  6. Before drafting TODO edits, present 2-3 plan shapes when there is a real grouping, sequencing, or write-location decision to make.

- Recommend one option based on the codebase review. - Keep the options at the bottom of the message under **Options**, with selector-prefixed labels such as A., B., C..

  1. Draft proposed edits (grouped by file):

- PRD/Blueprint deltas only if required - TODO step updates using the repo’s existing Step template - translate spec deltas into implementation deltas instead of restating product intent - for each new or revised execution step, include exact boundaries, interface or contract changes, sequencing notes, and validation evidence - add Implementation notes when the step would otherwise force the implementer to make decisions - split oversized mixed-surface work into dependency-ordered steps

  1. Ask: Approve and apply these changes?
  2. After applying, suggest implementing the next step via $cdd-implement-todo.

平台分布

Codex

37.29%
按下载量换算28

Claude

30.97%
按下载量换算23

Cursor

18.05%
按下载量换算14

Gemini CLI

9.73%
按下载量换算7

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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