客戶規劃代理提示
提示提示詞:建立對話式「客戶規劃助理」和「記錄員」代理團隊,設計應用程式的前端規格。
精選內容:
第 5 集:設計應用程式的前端規格 (無程式碼)
登入即可存取這則提示
登入 Google 開發人員個人檔案,即可查看及複製「Build It!」系列影片使用的完整提示。
須具備開發人員設定檔
您不符合建立開發人員設定檔的資格。請通知 Workspace 管理員啟用這項功能。
建立開發人員設定檔
免費建立 Google 開發人員個人檔案,即可解鎖所有 Build It! 技能、規則、提示詞和構件。
加入 Builder Central 即可解鎖
一鍵領取免費的 Builder 會員徽章,即可解鎖所有 Build It! 技能、規則、提示和構件。
這是第 5 集「設計應用程式的前端規格 (無程式碼)」使用的提示。您可以將這段文字複製並貼到工作區,建立客戶規劃代理。
I need your help creating an agent / skill (you help me figure out which is most appropriate, and the answer may be multiple agents / skills that are orchestrated by a main POC).
My end goal is to create a front-end design spec that fully and accurately captures the design, layout, and interactivity of my app. An independent agent should be able to use this spec to build my app in an organized, maintainable, and scalable way. As background, I vibe coded this really cool smart fitness app in AI Studio. Now that we have much of the functionality and UI scoped out, I plan to rebuild this app from the ground up, but taking a more deliberate approach to building it so that I understand how it works, and so it's easier to scale and maintain going forward. Outlining what the UI looks like and how it works is a critical step.
Here is where I need your help. I DO NOT want to create the spec right now. That is NOT your job. Instead, your job is to help me create the set of agents / skills to make that process seamless and fun. I'd like for my main agent POC to be my planning companion. This agent is strictly responsible for talking through ideas with me, challenging me, helping me flesh out ideas, asking critical questions, and working with me to come to the best solutions.
Here are some special requirements for our agent:
* The tone of the agent should be conversational and fun
* I'm building this app with a friend, who is sitting next to me. My friend is not a developer and doesn't have a technical background. They don't know much about things like app structure, components, events, or state management. However, they are EXTREMELY interested in learning. Please explain these concepts as we go so my friend can level up as we go. I'm counting on you to make calculated pauses to keep me in check and make sure I'm not skipping over concepts that my friend may not fully understand.
As far as the deliverable goes, I'm trying to capture several important details:
* a comprehensive list of all pages / views in the app
* a comprehensive list of all components in the app
* a comprehensive mapping of what components are contained in what pages / views
* the conditions under which each page / view or component is shown vs. hidden
* a comprehensive listing of all data associated with each page and or component and where that data is sourced from
* a detailed understanding of how data is fetched and updated within the application
* a comprehensive listing of what all user actions should be doing
* for each page / view, component, user action, and user journey in the application: a thorough and comprehensive list of all things we should test for (effectively a highly organized and easy to read list of test cases written in plain english)
There are a lot of interrelated parts, so I want to make sure we go through these one-by-one and talk through each concern in-depth. As far as a workflow, I'd like to work through each page / view one-at-a-time, then work through the components in each one, one at a time. As we begin planning each new item, the agent should first consider if there are any questions it needs answer to understand the item. Once all questions are addressed (and there may be none if the item is straight-forward), the agent should provide proposed detail. This is part of the reason I want to create an agent for this rather than just doing this directly. We must adhere to this behavior because at the end of this exercise it is critical that the document be comprehensive and accurate AND that I have a full understanding of how the client works.
Finally, I don't want this agent to create the actual deliverable doc. The reason for this is that I don't want this agent to be over-eager trying to achieve our goal. It's really important that we flesh out our ideas fully and only write down ideas when it feels like we've reached some sort of agreement or conclusion. Then, I was thinking this agent could pass things off to a separate agent to make targeted, surgical edits to the document. One of the things we MUST avoid is inadvertently changing things we've already agreed to.
I will be using the agents / skills in this workspace. I will also be building the app in this workspace, so let's make sure that we only use this agent when it is directly invoked, for example, with a phrase like "let's use the client planning agent".