20 个原创小情境，看看你如何与 AI 搭档。 四条习惯轴、八个友好原型，以及一个可以马上试的小实验。
打开应用 / Open app · 返回互动实验室 / Studio · 方法与边界 / Method
这是关于自述协作习惯的反思游戏，不是经过心理测量验证的人格测验，也不诊断、预测能力或推断人口特征。题目、报告和原型坐标是原创的产品规则；应用不调用大模型。
- 中文默认，完整英文切换；20 题、每轴 5 题，约 3—5 分钟。
- 回到上一题、刷新续答、结果页逐题回改；没有“正确答案”。
- 公开反向题、等权评分、四轴位置和八个手工原型；精确并列完整保留。
- 结果含优势线索、适当调整和具体流程小实验；每种路线都保留最终人类审核。
- 可编辑协作提示词，检查后自愿复制到自己选择的 AI 对话；本站不代发。
- 1080 × 1620 PNG 分享卡，在本机生成，不含逐题原始答案，也不会自动发布。
- JSON 备份包含原始答案。完整校验和确认后才替换；保存失败则保持现有进度。损坏的旧记录保留原文，供手动导出。
- 不需要账号、依赖包、构建步骤、分析 SDK 或外部 API。
查看分享卡示例（明确标注人工合成的中间选项；非真实个人结果）。
An original, bilingual reflection game about self-reported AI collaboration habits. Twenty scenarios map onto four axes and eight friendly narrative prototypes. Each report offers useful workflow clues and a concrete small experiment.
This is not a psychometrically validated personality test. It does not diagnose, predict ability, infer demographic traits, or call an LLM. Positions are not probabilities, percentiles, or grades.
Answers can be revisited, edited and resumed after reload. Scoring and hand-authored prototype coordinates are visible in the app. Exact nearest-prototype ties stay ties. JSON imports are validated before confirmation and installed only after a successful storage write. A damaged saved record is preserved rather than silently overwritten.
Take an aggregate-only 1080 × 1620 share PNG, copy result text, or edit a structured prompt and voluntarily paste it into an AI chat. Raw answers are included only in the JSON backup. No account, build, third-party package, analytics SDK, or external API is required.
Serve this directory with any static HTTP server, then open its local URL. Direct file:// opening can block ES modules. For example:
python -m http.server 8037Use Node.js 20+ for the dependency-free checks:
npm run check
npm testGitHub Pages can serve the repository root directly; no build output is needed. The CI workflow runs syntax checks and the core tests.
| File | Purpose / 用途 |
|---|---|
data.mjs |
Original bilingual questions, axis labels, reports and prototype coordinates / 原创双语内容 |
core.mjs |
Pure scoring, ties, strict backup validation, text/prompt/card preparation / 纯规则核心 |
app.mjs |
Accessible quiz, local persistence, import confirmation and downloads / 页面交互 |
shared.css, shared.mjs |
Interactive Studio styles and client-side copy/card helpers / 实验室共用组件 |
tests/core.test.mjs |
Reverse scoring, ties, malformed imports, resume and aggregate-only output / 核心测试 |
docs/METHOD.md |
Exact rules, versions, privacy and interpretation / 方法说明 |
The content version is 1.0.0 and backup schema is 1. Incompatible versions are rejected; no automatic migration is claimed. Storage belongs to the current browser and origin. Clearing site data, private browsing, browser policies or changing the deployment address can affect resumption; export a backup when you want to keep answers.
rubynor/bigfive-web was inspected as a public example of a localized questionnaire-to-report product flow. Its questions, text, code, five-factor constructs and scoring were not reused. This project uses original AI collaboration scenarios and explicitly unvalidated reflection rules.
MIT · JosephSun1854