Sdlc Intent logo

Sdlc Intent

Community
TsCarpe
sdlc-intent

Digests and health-checks product requirement documents (PRD). Stage 1 produces a structured digest (roles, concepts, feature map, flow diagrams, state machines) to build shared understanding; stage 2 audits the requirement against a seven-layer defect taxonomy (six in-document consistency layers plus an external-reference layer for missing-content defects) and produces a graded issue report plus a PM clarification checklist. Use when the user shares a requirement/PRD document (Feishu/wiki link or local file) and wants to understand it, prepare for design, or find problems before development. 触发词:需求文档、需求梳理、需求体检、需求评审、产品需求、帮我理解这个需求、PRD digest、requirement audit.

Overview

PublisherTsCarpe
Repositoryclaude-sdlc-skills
Skill namesdlc-intent
Stars
65
Forks
0
Bundled files
25
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 25 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by TsCarpe on GitHub. Read the source before you install it.

Installation

Install the Sdlc Intent AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/TsCarpe/claude-sdlc-skills/tree/main/skills/sdlc-intent
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/TsCarpe/claude-sdlc-skills.git /tmp/claude-sdlc-skills
mkdir -p .claude/skills
cp -r /tmp/claude-sdlc-skills/skills/sdlc-intent .claude/skills/sdlc-intent
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Sdlc Intent in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Sdlc Intent on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Sdlc Intent is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

需求接收助手(sdlc-intent)

两段式:第一段梳理建立对需求的共识理解,第二段体检尽早暴露需求缺陷。梳理是体检的前提——先与用户对齐理解,再找问题。

前置依赖:python3(parse_doc / scan_refs / resolve_anchors,纯标准库);飞书输入用 lark-cli(未安装/未登录或权限不足时,按「输入读取」降级为本地文件或文本)。

选择工作流

用户说「评审 / 梳理 / 找问题」时先确认宾语:对象是需求文档才进本 skill;是设计/用例走 sdlc-gate,是技术方案走 sdlc-doubt,是上线前检查走 sdlc-release-check。

按用户意图选择分支:

  • 用户说「梳理需求 / digest / 理解需求」,或首次给出需求文档 → 只跑【第一段:需求梳理】,跑完暂停
  • 用户说「体检 / 需求评审 / 给需求找问题 / review」→ 跑【第二段:需求体检】;若已存在梳理文档(sdlc/*/intake/digest-*.md)优先复用,不存在则提示先梳理
  • 复用梳理文档前强制版本校验:体检本身要重新读取原文,读取后将本次原文的版本/修订时间与梳理文档头「版本」字段比对——一致 → 直接复用;不一致 → 告知用户「原文已从 vX 更新至 vY,梳理文档基于旧版」,建议重梳理(用户明确说沿用才沿用);原文无版本信息时,向用户展示梳理文档的梳理日期,请其确认原文是否有更新
  • 用户只给文档未指明 → 两段连跑:梳理完成后暂停,向用户展示摘要并提示「请确认或补充业务理解,完成后说“继续体检”」,得到确认再跑体检
  • 用户要求「开始开发 / 拆任务 / 写代码」时:先走 sdlc-gate 的开发请求入口;未过 gate 先停,见「下一步(流程衔接)」

输入读取

  • 飞书链接(feishu.cn / doubao.com 的 /docx/、/wiki/):lark-cli docs +fetch --doc "<URL>" --doc-format markdown(权限不足时向用户说明,不要反复重试)
  • 原文预处理(仅飞书输入;正文与评论读取后执行一次):拉取块级 XML(lark-cli docs +fetch --doc "<URL>" --doc-format xml --detail with-ids),运行 scripts/parse_doc.py 产出 sdlc/<需求名>/intake/doc-<YYYYMMDD>.json 与 media/ 目录(输入提取、命令行、块模型、条款级出处、锚点附加与失败回退详见 references/doc-pipeline.md);parse_doc 内置内容守恒门(conservation-v1,exit 2 = 产物与原文不一致、未落盘、不得回传),doc.json 顶层带 guard 供平台落库终检;doc.json 存在时,梳理/体检出处尽量写到条款级(「详述N·<小节名>」/「详述N·<序数>」);解析失败不阻塞梳理(平台回退文本定位)
  • 飞书链接在读取正文后必须同步读取评论区:lark-cli drive +list-comments --url "<URL>" --solved-status all --comment-scope all --need-relation --format json(权限不足时同样只向用户说明一次、不重试,正文流程照常继续)
  • 评论区是正文的补充信息源,正文外的关键决策/变更说明/答疑常在评论区;评论结论与正文不一致时以最新评论为准。本地文件输入或评论数为 0 时记「无评论」,权限不足记「未读取(原因)」,均如实写入梳理文档头「评论区」字段
  • 本地 markdown/text 文件:直接 Read
  • 文档内嵌表格已随正文返回;若正文出现 <sheet>/<bitable> 引用且与核心流程相关,用 lark-sheets/lark-base 下钻读取(有权限时)
  • 正文出现内嵌附件块(<figure ...><source ... token="...">)且为原型/演示文件(html 等)时,必须下载读取:lark-cli docs +media-download --token "<token>" --output <本地路径>——markdown 拉取后附件只剩一个自闭合标签,不下载则原型功能面信息完全缺失(原型有而正文无的功能是体检高价值线索)。html 原型体量大(可达数百 KB),只提取结构(页面/导航/弹窗/字段/操作)入上下文,不全量读取。下载失败记「未读取(原因)」写入梳理文档头「原型附件」字段,同评论区口径不重试。原型结构仅供体检阶段交叉核对,不并入梳理功能地图——梳理仍忠实正文+评论区
  • 原文内容(正文 + 评论区 + 原型附件)在后续体检阶段还要反复引用,梳理时先在本地留存原始内容(如 /tmp/ 下临时文件;评论 JSON、正文 markdown、原型文件分开存放)
  • 变更单增量梳理:变更单 confirm 后的受影响章节局部重检规范见 references/incremental-guide.md(增量包结构、作废清单、anchor 越界纪律;平台侧执行,sdlc-sync 推送时消费)

第一段:需求梳理

复制此 checklist 跟踪进度:

梳理进度:
- [ ] Step 1: 通读原文,提取角色与术语
- [ ] Step 2: 按 references/digest-template.md 生成 8 区块梳理文档
- [ ] Step 3: 落盘 sdlc/<需求名>/intake/digest-<YYYYMMDD>.md
- [ ] Step 4: 终端展示摘要(含风险分级提议)+ 暂停等待用户确认与分级拍板

Step 2 是核心:读取 references/digest-template.md,严格按模板 8 区块产出,功能地图、流程图、状态机、规则表必须图表化(markdown 表格 + mermaid)。

梳理纪律(违反即返工):

  • 忠实转述原文,不评价、不脑补、不优化——评价是体检阶段的事
  • 原文缺失的流转/触发/字段,用 🤔 标注,禁止自行补全
  • 对模糊内容的个人理解,逐条记入「假设清单」区块(编号 + 原文位置 + 按 X 理解),并标注类型(推断 = 有原文推导链 / 假设 = 无依据拍板)
  • §7 关键规则逐条编 FR 号、状态机图每条流转须有原文出处——编号与出处纪律详见 digest-template 标注约定(FR 编号稳定性、ID 命名空间、评论区判定口径均以模板为权威)

Step 3 落盘路径:当前项目 sdlc/<需求名>/intake/ 目录(不存在则创建),文件名 digest-<YYYYMMDD>.md。

Step 4 暂停:展示各区块摘要(尤其 🤔 集中处和假设清单)与风险分级提议——按 references/risk-tiering.md 评分卡给出 A/B/C 提议+影响面命中依据(拿不准从高,降级需附可核对证据),明确说「请确认或补充业务理解,并对风险分级拍板,完成后说“继续体检”」;拍板后将 digest 头「风险分级」的确认改为日期,然后停止等待。

第二段:需求体检

分级路由(Step 1 时判读):读 digest 头「风险分级」字段——A 级全阶段跑(阶段0+阶段1+七层);B/C 级跳过阶段1(视角推导),主会话顺序跑。无该字段(存量文档)= 未分级 = 按 A 级处理并提示补标。判定结果写入体检报告执行摘要首句。分级判定与回写规则见 references/risk-tiering.md。

复制此 checklist 跟踪进度:

体检进度:
- [ ] Step 1: 准备输入(原文 + 梳理文档,含用户补充)+ 分级路由(读 digest 头风险分级,缺失按 A 级)
- [ ] Step 2: 阶段0 参照系构造——按 references/ref-tables.md 产 refs-<日期>.md(可先跑 scripts/scan_refs.py --mode seeds 供料,构表后 --mode verify 校验)
- [ ] Step 3: 阶段1 视角推导(仅 A 级)——按 references/perspectives.md 扇出实现者/测试者推导,产出 D2 候选
- [ ] Step 4: 按 references/dimensions.md 七层维度逐层检查(第 7 层汇流 D1/D2/D3 候选做语义判定)
- [ ] Step 5: 按 references/report-template.md 产出体检报告 + PM 确认清单(命中带来源标注)
- [ ] Step 6: 终端摘要(各级问题数 + 阻断项列表)
- [ ] Step 7: 交互收尾(用户逐条回应后回写梳理文档)

Step 2(阶段0·参照系) 读取 references/ref-tables.md:从梳理产物构造四张参照表(数据字典/指标字典/平行结构矩阵/引用图,按触发特征裁剪),落盘 sdlc/<需求名>/intake/refs-<YYYYMMDD>.md。表中空格、断链、⚠️冲突格即缺失类候选,交给 Step 4 做语义判定;某张表构造不出来本身也是信号,如实记录。可选机械辅助:python3 scripts/scan_refs.py --mode seeds --doc <原文|doc.json>(构表前产断链/并列章节 seed)、--mode verify --doc <原文> --refs <refs.md>(构表后产字段出现位置清单——有录无消的证据义务、机械候选与计数核对);脚本不可用或输入不含时跳过,不阻塞。

Step 3(阶段1·视角推导,仅 A 级) 读取 references/perspectives.md:以实现者/测试者身份执行推导任务(产出数据流草图/断言草稿,缺陷是推导副产品),用 Agent 工具扇出全新上下文子代理(喂料纪律:只喂任务卡+模块切片+相关表行,不喂全文);候选按收敛纪律(反向锚点/不与 D1 重复/待裁决只追问)回主会话,标 D2。

Step 4(七层核对) 读取 references/dimensions.md,按层序检查:概念层最先(术语表是后续所有检查的基准),第 7 层外部参照层最后——汇流阶段0 候选(D1)、阶段1 候选(D2)与本层推导(D3)做语义判定与统一分级。该文件定义了每个检查项的命中标准与严重度;规则类/决策类命中按其「决策类型标注」规则在报告类型列标 C 约束 / H 假设(初判带问号)。

体检纪律:

  • 每条命中必须附原文定位(章节名 + 关键引文片段),禁止脱离原文泛泛而谈;缺失类无引文可引时按「全文未见 X」问式(见 report-template 撰写规则)
  • 七层维度是启发式锚点而非封闭清单:发现维度外的问题也应记录,标注维度为「补充」
  • 严重度从严不从宽:拿不准阻断还是严重时,标高一级并在描述中说明(第 7 层例外见 dimensions.md)
  • 冲突不裁决:构表或层间检查撞到的口径冲突一律记命中交 PM 裁决,禁止静默择一

Step 5 读取 references/report-template.md,产出两个文件到 sdlc/<需求名>/intake/:

  • audit-<YYYYMMDD>.md(五章结构:执行摘要/阻断速览/分层明细/主题关联/状态汇总——唯一的成段文字在执行摘要,其余表格化)
  • pm-checklist-<YYYYMMDD>.md(可直接整篇复制发给产品经理的确认清单)

两个文件产出后,必须按 report-template.md 的「终检」清单核对(统计表重数 / 跨文件计数一致 / 主题组关联核对),通过后才进入 Step 6。

Step 7 交互收尾:引导用户对每条问题回应「问 PM」/「按假设 X 处理」/「忽略」;有结论后将其回写至梳理文档的假设清单,形成共识版。用户不想逐条过时直接结束,不强制。

下一步(流程衔接)

  • A/B 级:intake 确认后 → /sdlc-design <需求名> 与 /sdlc-test cases <需求名> 并行独立产出 → /sdlc-gate <需求名>;gate 放行前不写业务代码。
  • C 级:/sdlc-design <需求名> 产一页 design/plan.md → /sdlc-gate <需求名> C 档主会话自查 → 人工确认放行后开发。
  • 用户直接要求开发:先走 sdlc-gate 的开发请求入口;未过 gate 先停,不得把梳理/体检完成当作放行。

边界

  • 本 skill 只处理需求文档本身,不读代码库、不做技术设计(设计属后续任务)
  • 体检只审需求本身的质量,不评估技术方案的合理性(方案合理性属后续技术设计评审范围)
  • 一次处理一份需求文档;多份文档时请用户明确主文档
  • 产出语言跟随需求文档语言(中文文档产出中文)

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Sdlc Intent AI skill do?

Digests and health-checks product requirement documents (PRD). Stage 1 produces a structured digest (roles, concepts, feature map, flow diagrams, state machines) to build shared understanding; stage 2 audits the requirement against a seven-layer defect taxonomy (six in-document consistency layers plus an external-reference layer for missing-content defects) and produces a graded issue report plus a PM clarification checklist. Use when the user shares a requirement/PRD document (Feishu/wiki link or local file) and wants to understand it, prepare for design, or find problems before developmen...

Why use Sdlc Intent on TypingMind?

Because you install it once and use it with any model. Sdlc Intent is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Sdlc Intent in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/TsCarpe/claude-sdlc-skills/tree/main/skills/sdlc-intent. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Sdlc Intent?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Sdlc Intent?

As many as you like. As long as a model supports skills, you can use Sdlc Intent with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Sdlc Intent AI skill free?

Yes. It is published on GitHub by TsCarpe under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇