把纷杂的输入,锚定为清晰的判断与可执行的共识。
From scattered context to grounded requirements.
Requirement Anchor is a reusable Codex skill for turning meetings, conversations, screenshots, files, and unfinished ideas into requirements that people can understand, build, and verify.
这个 skill 起源于一套日常工作中的个人实践:面对会议记录、聊天片段和零散想法,先辨认真正的问题,再厘清已经确认的事实,区分讨论中的方案与最终结论,最后将理解沉淀为可以执行和验收的需求。
It grew from an everyday working practice: understanding the real problem behind messy context, separating facts from assumptions, and shaping conversations into requirements that teams can build and validate.
它不是某个公司的专属流程,也不是一套必须照搬的模板。你可以直接使用,也可以按照自己的团队、项目和思考方式重新塑造它。
It is not tied to any company or rigid process. Use it as a starting point, adapt it to your team and projects, and make it your own.
需求往往不是从一张白纸开始,而是从一场没有结论的会议、一段不完整的对话,或一个还没有被说清楚的想法开始。
Requirements rarely begin as a blank page. They begin as an unfinished conversation, a noisy meeting, or an idea that has not yet found its shape.
Requirement Anchor helps you:
- 从零散材料中找到真正需要解决的问题;
- 区分事实、结论、假设、待确认事项和后续规划;
- 识别会议转录中的错词、断句和术语问题;
- 根据实际复杂度组织需求,而不是机械套用模板;
- 将需求写到产品、开发和测试都能理解和验证的程度。
It helps you:
- find the real problem behind scattered material;
- separate facts, decisions, assumptions, open questions, and future plans;
- interpret noisy transcripts and correct ambiguous terminology through context;
- choose an appropriate form instead of forcing every request into one template;
- produce requirements that product, engineering, and QA can align on and verify.
- 会议记录、聊天记录、截图或自由描述转需求;
- 产品、运营、后台和技术优化需求;
- 已有需求的评审、重写和补充;
- 复杂讨论中的结论提取与范围梳理;
- 需要时补充开发、测试和实现规划。
先理解,再成文。
Understand first, write second.
不把猜测写成事实,不把建议写成结论,不把未来规划混进当前交付范围。需求的价值,不只在于写得完整,更在于让不同角色对同一件事形成可靠的理解。
Do not turn guesses into facts, suggestions into decisions, or future plans into current scope. A good requirement is not merely complete; it creates a reliable shared understanding.
支持中文和英文。用户可以直接指定输出语言;如果没有指定,skill 会跟随输入材料的主要语言。
Supports both Chinese and English. Specify the output language when needed; otherwise, the skill follows the primary language of the input.
你可以修改 SKILL.md 和 references/ 下的内容,加入自己的需求字段、团队术语、评审规则和输出偏好。它更像一枚可以不断校准的锚,而不是一块固定不变的模板。
You can customize SKILL.md and the files under references/ with your own fields, terminology, review rules, and writing preferences. Think of it as an anchor you can keep calibrating—not a template carved in stone.
Released under the MIT License.