[loopx-agent]

Why

The packaged Goal workspace smoke asks the view to show an unknown blocking
count when it can see only a bounded sample. The status fixture now supplies
21 classified rows in the top-level queue and three in the overview. In the
old mixed case the UI correctly uses the 21 rows and shows one blocker, so the
smoke fails before it tests partial coverage.

Change

Use the canonical producer's IDs to supply three rows for a genuinely partial
mixed queue. Exercise overview fallback and absent totals separately. Retain a
mixed complete case with the full canonical queue and its bounded overview.
The existing exact count assertions now distinguish one blocker and 20 optional
items from a genuinely unknown blocking total.

Evidence

On the same installed Chat bundle and synthetic Goal data, the original
scenario fails with one blocker where it expects an unknown total. The
corrected scenario passes in Chinese and English, including desktop and narrow
layouts, with the strip and Overview reading the same qualified count. It also
passes the existing stop readback and no-write assertions.

[loopx-agent]

Reviewer: model_agent, gpt-6.1-sol, OpenAI, runtime_reported, reasoning_effort=max

Approval conclusion (author-owned PR; GitHub blocks formal self-approval)

没有阻塞发现。

动机 / Motivation

运行打包 Workspace 浏览器检查的维护者，以及读取部分 Goal 注意事项的用户。 同一隔离安装的 v2.2.0 wheel、Chromium、合成状态和命令。原用例在 mixed 模式读到实际阻塞数 1，却期望 —。当前用例只留下三个原 Todo ID，读到 —。 中文 1512px、英文 390px 各验证 7 种状态。完整 mixed 为 1 个阻塞和 20 个可选，只有可选为 0 个阻塞。完全空样本不造卡，缺总数时仍未知。状态条和 Overview 相符，未发起 Turn 或写入。 改动限定于浏览器 fixture 与断言。 只验收本 smoke 和实际消费链。

改动思路 / Approach

复用 canonical producer 的三个 Todo ID 构造部分覆盖。把概览 fallback、缺少总数、空样本和完整顶层分开测试。完整 mixed 用同一份 21 条来源，验证 1 个阻塞和 20 个可选。

依据 94b91e4 的 Smoke review checklist。

• Representative path 真实状态 producer 经安装包的 parser 和 adapter 到达两个 UI 消费者。
• Independent oracle 用实际分类和覆盖范围推导计数，分别断言 1、0、21 和未知。

具体改动 / Changes

一个已有 smoke 文件 +14/-11，减少一次 producer 调用。原 stop readback 通过。 Fixture and assertions

对主干的风险 / Risks

Quick premerge 的 4 项直接检查和 3 项选中检查通过。 目标分支和当前提交的组合无冲突，前端源码和打包入口一致。

我的整体评价 / Verdict

当前 head 修复了可复现的测试误报，保留了原停止及权限检查。

English verdict: APPROVE this exact head. The packaged smoke now distinguishes partial coverage from a fully classified Goal queue.