[BigDataDZ]

The pending-report lifecycle regression still requires a generated title-only masthead summary, but current HTML intentionally omits that duplicate when authored highlights are present. It therefore reports a failure even though the actual conclusion, risk and next-action content is rendered. Native Windows also fails four adjacent lifecycle cases because their test-side file/subprocess reads decode UTF-8 report data with the locale.

Verify authored titles and substantive paragraphs in report articles, excluding script/style content, instead of requiring redundant summary copy. Keep exact-replay, delivery Todo, normalized document, digests and lineage assertions. Make the two editorial-request reads and real CLI capture explicitly UTF-8. No renderer, UI or product lifecycle behavior changes.

Fixes #7437.

Validation at signed head 3854bcda1, based on main 218aa087b:

• Original focused report case fails on the retired masthead string. On clean Windows it first hits the separate missing IANA database; local test prerequisites include tzdata, whose Windows installation fix is proposed independently in fix(install): supply IANA timezone data on Windows #7454.
• Final complete pending-intent + HTML modules: 94 passed on native Windows / CPython 3.12.10 with PYTHONUTF8=0.
• Before the four test IO fixes, the same modules had 90 passed / 4 failed; the four failures reproduce on unchanged main (4 failed / 40 deselected, with the original decode warnings). The final run fixes those test reads without changing the real CLI output encoding or authority behavior.
• A process-local mutation removes only the authored overview paragraph from actual renderer output. The revised focused lifecycle test fails on the missing article content (1 expected failure). Original renderer source bytes remain unchanged; this is a semantic negative control, not a product failure.
• Changed-file Ruff and git diff --check pass. Only the existing test module changes; private probes and logs are excluded.

Author declaration: authored with OpenAI Codex under contributor direction. This restores durable end-to-end report-content and native readback coverage under the existing report owners. Frontend/Lark/CLI entrypoints and first-screen rendering are unchanged. The bounded future-facing pass keeps encoding explicit at the actual test IO boundaries and avoids a new parser module or fixture framework. No model/benchmark job, external delivery, active Goal mutation, system upgrade or self-merge. Independent review and current-head CI remain pending; maintainer merge required.

[loopx-agent]

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

动机 / Motivation

维护报告生成与重放回归的贡献者。 旧测试要求已移除的重复页眉摘要，实际正文存在时仍会报失败；部分测试读写还依赖本地编码。 从实际HTML的article中验证作者的结论、风险、下一步及正文，重放和交付身份断言保留。 仅修改已有测试；不改变renderer、报告内容、调度、外部发送授权或用户界面。 当前实际报告内容仍完整，测试应验证读者能读到的作者正文，而不能把已移除的重复页眉当作产品义务。完整精确head为3854bcda12163263eb7da7580d7cc8b0a27583bc。

改动思路 / Approach

在原生命周期测试内解析可见article文本，并在实际测试IO处指定UTF-8；复用现有产品pipeline与独立HTML模块。 它继续调用真实consumer、renderer与文件写入，不复制生成逻辑，也不修改产品使旧字符串重新出现。原有重放、交付Todo、规范化文档、摘要和lineage检查保留。article以外的copy数据、脚本与页面页眉不再能冒充作者正文；这里只解析当前既有article结构，不宣称另建了通用浏览器可见性引擎。

具体改动 / Changes

规格docs/reference/protocols/periodic-report-v0.md，spec_revision=59c8bbf791bca9d3d1dd996629a1610c0679b23e：固定文本。AUTHORED-CONTENT implemented：六项作者标题/正文均从实际article读回；REPLAY-AUTHORITY implemented：原精确重放、delivery Todo、摘要和无外部写断言保留；UTF8-READBACK implemented：两个editorial request读取和两次实际CLI捕获明确指定UTF-8。当前本地完整验证与作者原Windows证据分别归属，不互相冒充。

关键代码讲解

• ArticleText与生命周期断言 将实际保存的HTML限定到article数据，要求结论、风险、下一步的标题与实质段落共六项；不是从当前输出生成期望值。
• 同文件两个editorial request读取及最终real CLI readback固定UTF-8，保留原生产stdout与文件契约。没有全局更改locale，也没有去掉Unicode输入或重复调用。
• 最强负对照仅在进程内将真实renderer输出的article overview段落移除，同时保留隐藏源数据并重新计算合法artifact内容摘要；既有新断言仍明确失败于assert authored in visible，说明它保护了可读正文而非只匹配嵌入JSON。产品源码字节未变。

对主干的风险 / Risks

当前精确提交两个完整模块94通过，零skip；移除实际article中一段作者正文的进程内变异导致原新断言失败。五项premerge直接检查、Ruff、公开边界和CQR通过。 首次本地命令写错HTML测试目录，未运行测试并保留为失败；纠正为实际presentation模块后94项通过，没有覆盖早先记录。变异的1项失败是预期反例，不算产品绿测。未查询、轮询或等待GitHub CI。局部parser依赖当前renderer的article边界，完整presentation模块提供邻接保障；正式页面渲染与移动端仍由发布浏览器资格承担。本次diff没有改变任何渲染输出，因此视觉检查not_applicable，未验证视觉范围none：没有视觉变化。作者近期八项提案已比较，本次与时区依赖、注册编码、调度夹具分别属于不同边界；没有重复smoke或贡献限制依据。

我的整体评价 / Verdict

APPROVE，goal_achieved，限于修复既有内容与读回回归。long_horizon preserved：重放、身份与权限语义保留；user_experience preserved：实际报告及操作不变，没有模型或长期净收益主张。bounded refactor applied：IO编码留在真正读取处，正文parser留在唯一需要它的原测试中。此宿主为macOS，未重新执行原生Windows；作者Windows结果保留来源。整版App和发布渠道资格另行验证。 合并仍需即时readiness、当前CQR与维护者授权。

English verdict: APPROVE — 3854bcda12163263eb7da7580d7cc8b0a27583bc. All94 full lifecycle/presentation tests pass,and an article-only missing-paragraph mutation fails at the intended assertion. Existing replay,digests and delivery authority remain unchanged; native Windows and whole-release qualification stay separate.