DispatchDock v0.1.7

改动 / Changes

• 审美定稿要能看、能点，不能只写文字。 dual-role-workflow 新增总原则：视觉、交互、动效、文案这类用户直接感受到的东西，设计负责人的定稿必须是界面稿、可交互原型、样片这样的产物，文字只记录"为什么"。网页 / 应用类项目改为在可交互原型里定稿视觉、交互和动效，Codex 照原型实现，时长、曲线、先后顺序和原型一致，原型没覆盖到的留 TODO(design)。动效和交互的时序一律算"不能客观验收"，不派给 Codex；验收时把实现和原型并排对照，不能只看截图和断言。角色文档模板同步补上。
  Aesthetic decisions must be finalized as something you can see and click, not just text. dual-role-workflow adds a general rule: for anything the user directly experiences (visuals, interaction, motion, copy), the design lead's final version must be an artifact such as a mockup, an interactive prototype or a sample clip; text only records the "why". For web / app projects, visuals, interaction and motion are finalized in an interactive prototype; Codex implements it with durations, easing and ordering matching the prototype, leaving TODO(design) where the prototype doesn't cover. Motion and interaction timing always count as "not objectively verifiable" and are not dispatched to Codex; review compares the implementation side by side with the prototype instead of relying on screenshots and assertions. The role doc templates are updated accordingly.

装了软件的：在"连接"页更新技能。只装技能的：用新的压缩包重新运行 node installer/install.mjs。两种都会先备份。
App users: update the skills on the Connections page. Skills-only users: rerun node installer/install.mjs from the new zip. Both back up first.

SHA-256

fc9e70d63b9ebdcaab567893dac0387e4d0907342b6f139b431cb27fe8dd5a32  DispatchDock-Setup-0.1.7-x64.exe
6578062f6ab64c75521c6738037261f19dddb7c4219a1cf067169a37a06514a1  dispatchdock-skills-0.1.7.zip