Skip to main content

beta-fleet:在发布前运行内测

在发布前让五个不同角色的首次用户子代理测试应用,修复遇到的问题,重新运行新用户以验证修复并通过 HTML 报告结果。

来源信息

仓库
seanblundin/beta-fleet
最近来源活动
2026年10月19日 16:43
检测到的 SKILL.md 语言
英语
星标
0
Forks
0

用途

当用户输入 /beta-fleet 或要求像新用户一样对应用进行内测时使用。

操作说明

  1. 启动应用并定义首次成功
  2. 在浏览器标签页中启动五个代表不同用户画像的子代理
  3. 合并并一次性修复问题
  4. 重新运行新的子代理以证明已修复
  5. 通过 HTML 页面汇报结果

引用来源

安装方式

默认使用会先检查来源的 Prompt;你也可以切换为直接命令,或下载本地副本。

检查来源文件

决定是否安装前,请先阅读 SKILL.md,以及 SkillsMP 当前展示的配套文件。

正在显示 SKILL.md

SKILL.md
来源说明 · 只读预览
name
beta-fleet
description
Send five first-time users (subagents in real browser tabs) through an app before launch, fix what they hit, rerun fresh users to prove it, and report back as one clear HTML page. Use when the user types /beta-fleet or asks to beta test an app like a new user.
*Last edited: 2026-10-19* # Beta fleet Built on [this tweet](https://x.com/RoundtableSpace/status/2111717667228770611): your first real users won't tell you why they left, but these five will. 1. Start the app (local dev server if there is one). Write one sentence for "first success": the first moment a new user gets real value. 2. Launch five subagents at once, each on Opus with high effort (`model: "opus"`, `effort: "high"`), each a different first-time user: 1. The impatient one: skims, clicks the first button, quits at the first error 2. The skeptic: reads pricing, terms and privacy before signing up 3. The mobile user: does everything on a 375px screen on slow 3G 4. The non-native English speaker: only follows what the UI literally says 5. The breaker: pastes junk into forms, hits back mid-flow, double-clicks everything Each opens the app in its own browser tab, goes from the landing page to first success, and screenshots every moment it gets confused, stuck or annoyed. On a local server, give each one its own host name (`http://<name>.localhost:<port>`) so their logins don't mix. If the browser tool can only drive one page at a time, run them one after another instead. 3. Merge everything into one list ranked by how many users hit it, and fix it all in one pass. Leave pricing, legal text and positioning to the user. 4. Rerun five fresh subagents that know nothing about the first run, to prove it's fixed. 5. Report back with one HTML page, opened for the user. Keep it short and plain. Use headers that say exactly what's below them ("Fixed", "Still broken", "Your call"), and put a before and after screenshot next to each issue.
在 GitHub 查看