在没有答案的情况下做一遍Do it once with no answer in front of you
每套模拟考都换了业务场景，但考点和真实 assessment 一一对应。这是检验「你是理解了还是背下来了」最直接的办法。Each mock swaps the business scenario but keeps a one-to-one match with what the real assessment tests. It is the most direct way to find out whether you understood it or memorised it.
它们不是源项目里的内容，也不是任何真实考题。 它们的作用是「换一个场景考同样的东西」。
两套模拟题的参考答案都在本机跑通过自己的测试： React 那套 5/5，Federation 那套 14/14。 starter 状态下的基线实测是 5 failed / 5 total 和 10 failed / 4 passed。They are not from the source projects and not real exam questions. Their job is to test the same things in a different scenario.
Both reference answers really were run here and passed their own tests: 5/5 for React, 14/14 for Federation. From the starter, the measured baselines are 5 failed / 5 total and 10 failed / 4 passed.
- 读完全部需求和评分标准，别急着看 starter。
- 在本机新建一个目录，把 starter 和测试抄进去。
- 按建议时长计时做完。中途不要打开讲解。
- 跑测试 + 做手动自检，按 rubric 给自己打分。
- 最后才打开讲解和参考答案，对照你漏掉了什么。
- Read every requirement and the rubric first, before you look at the starter.
- Make a fresh directory on your machine and copy the starter and the tests into it.
- Work to the suggested time limit. Do not open the walkthrough on the way.
- Run the tests and do the manual checks, then score yourself against the rubric.
- Only then open the walkthrough and the answer and see what you missed.
IT 支持工单看板。可以新建工单（标题 + 优先级）、关闭工单、改派（编辑）工单，还能按优先级筛选。业务场景换了，数据结构多了一个枚举字段和一个筛选状态，但底层要你会的东西和 Notes Manager 一模一样。
图书评论 subgraph。它既不拥有 Author 也不拥有 Book —— 两者都由 Catalog subgraph 提供，本服务只往它们身上挂 reviews 和 averageRating。Book 用的是复合 key（isbn + edition），这是真实项目里很常见、但比单字段 key 更容易写错的情况。