Kami V1.16.0

Good content deserves good paper.

Changelog

1. Long-doc page numbers: WeasyPrint 70 stopped resolving the counters and printed every one as 0, so the anchors are blocks again and --verify compares each numeral against the page its row links to.
2. Verified numbers only: every bullet and paragraph had to carry a figure, so it is verified numbers now and observable outcomes where no metric exists.
3. Diagram geometry checks: annotate nodes, edges, and label masks and --check reports misattached endpoints, overlaps, covered labels, and edges through a node.
4. Typography and revisions: body measure holds 40 to 70 characters, alignment is optical rather than mathematical, and variants render in the product's own frame.
5. Lighter intake: the page-count question is gone, document type and length come from your content, and the skill no longer restyles a website you already have.

更新日志

1. 长文目录页码：WeasyPrint 70 起页码在 flex 锚点上解析失败并全部渲染为 0，改回块级锚点后恢复，--verify 会比对页码与链接的目标页。
2. 数字要真：此前每条经历和每个段落都必须带一个数字，现在只用可核实的数字，没有指标时写明可观察的结果。
3. 架构图几何检查：标注节点、连线和标签遮罩后，--check 会报出端点未附着、节点重叠、标签被覆盖和连线穿过节点。
4. 排印与改稿：正文行长限定 40 到 70 个字符，对齐以视觉为准，行距映射固定档位，候选方案在产品自身背景中渲染。
5. 前置提问：不再先询问页数，文档类型与篇幅由内容推断，技能描述收窄后不再触发对既有官网的审计或改版。

Thanks

Issue reporters this cycle: @yuanyuanzhang0815 · @SinnLake.

Kami helps AI agents create clear, consistent professional documents. https://github.com/tw93/Kami