Convert meeting-note originals and Figma source material into a structured Markdown PRD. Extract facts from the meeting text when available; when recordings or notes are missing or incomplete, use Figma page structure, UI copy, frames, components, states, annotations, and prototype flows to complete the requirement draft. Export corresponding screenshots, associate design evidence with requirements, and write function descriptions as professional product manager PRD tables while clearly marking assumptions and pending confirmations.
Use this skill for requests such as:
Collect or locate the following inputs:
If meeting notes or recordings are missing/incomplete but Figma is available, continue in Figma-first mode: infer product scope, user paths, UI states, and interaction logic from the design, then mark inferred or unverifiable business rules as 基于设计稿推断,待确认. If Figma is missing too, continue with available material and mark unverifiable sections as 暂无 or 待补充 according to the output rules.
Extract the following facts from the meeting-note original:
Preserve important wording when it affects product logic. Do not expand vague statements into specific rules unless the design supports them.
When a Figma source file/link is present, load the figma skill before analysis. Use it to inspect file structure, pages, frames, components, variants, auto-layout hierarchy, text layers, annotations, prototypes, and exported assets as needed.
Use scripts/figma_extract.py as the built-in Figma recognition helper when an API token is available. The helper parses Figma URLs, extracts the file key and node id, calls the Figma REST API, and writes a compact JSON file containing file metadata, page/frame inventory, and target node data.
Run pattern:
python scripts/figma_extract.py "<figma-url>" "<output-json>"
Export screenshots for PRD illustrations with scripts/figma_export_images.py. Use the target node id from the Figma URL first; when multiple feature modules map to different frames, pass the mapped node ids explicitly.
python scripts/figma_export_images.py "<figma-url>" "<prd-assets-dir>"
python scripts/figma_export_images.py "<figma-url>" "<prd-assets-dir>" --ids "1086:4328,1086:4330" --format png --scale 2
Screenshot usage rule:
assets/figma/ or -assets/figma/ .
!入口页.
待补充截图 in the illustration column.
Authentication rule:
FIGMA_ACCESS_TOKEN from the environment or from a local .env file in the current working directory.
待补充/待校验.
Identify and record:
If Figma cannot be opened directly, use any provided screenshots/exports and clearly mark design-derived assumptions as based on exported material.
Build a requirement-evidence map before writing:
待确认 rather than choosing silently.
Use assets/prd-template.md as the output structure. Keep headings stable and output a clean .md document. When screenshots are exported, save them beside the PRD in a stable assets directory and reference them from the function-detail table with Markdown image syntax.
The PRD must include these sections in order:
# 需求标题
## 1. 需求背景和目标
## 2. 上线时间
## 3. 设计稿
## 4. 需求内容详细描述
## 5. 灰度放量方案
## 6. 埋点设计
Generate a concise requirement title:
【战绩玩法】+ 自总结标题
【战绩玩法】经典农场资料卡
【战绩玩法】需求整理.
Write background, problem to solve, value, and expected effect. Use 暂无 if not mentioned and cannot be inferred.
Include expected launch time and front-end/back-end workload only when mentioned. If all timing/workload information is absent, write 暂无.
Place the original Figma/design link directly. If multiple links are provided, list all. If only local files or exports exist, list file names/paths.
Follow the rules in references/prd-writing-rules.md. This section is the core output and must be written as a professional PM PRD, organized by user flow and functional module. For all major function descriptions, use a three-column Markdown table with 示意图、功能场景、需求详情; the illustration column should reference the exported Figma screenshot for the corresponding frame/node whenever available.
Translate experiment, gray-release, audience split, whitelist, staged rollout, or full-release strategy from the meeting notes. If no strategy is mentioned, write 暂无 or 默认全量发布(待确认) only when context clearly implies direct full release.
Reserve a placeholder only. Do not invent event names unless separately instructed. Output:
待补充:后续根据埋点规范补齐曝光、点击、转化及关键状态事件。
Before finalizing, verify:
待确认/不在本次范围.
待确认 item.
示意图 / 功能场景 / 需求详情 table format and include exported screenshots or 待补充截图.
共 1 个版本