[PR-4] M3 接入渲染门禁(paper_orchestra_pipeline.py:1843)

[justDance-everybody]

本任务属于 Epic #23。共享上下文(整体目标、Agent Execution Contract、Non-goals、回归基线)见 #23。
依赖:blocked by #25(需 PR-3 的四个检查先落地)。
代码定位按当前仓库版本核对;若行号漂移,以函数名 / 字段名 / 挂载点语义为准。

完成本 PR 需遵守 Epic #23 的 Agent Execution Contract 与 Test Execution Protocol:先写失败测试与反例夹具 → 最小实现 → 跑里程碑测试 → 跑回归基线。PR 描述需列出:changed files / tests added / tests run / before-after / non-goals。

---

目标

把 PR-3 的四个检查接进真渲染门禁:任一命中 → 禁止出稿。

Allowed changes

• agents/paper_orchestra_pipeline.py:1843 附近的渲染门禁接入逻辑
• M3 integration tests
• PR-4 fixtures

Forbidden changes

• 不修改 contracts/pipeline.py:626 require_submission_ready()
• 不绕开既有 latex_sanity_check
• 不只返回第一个 hit;多处命中必须全部列出
• 不改变现有 placeholder 命中行为

改动点(挂载点是 1843,不是 require_submission_ready)

• 正确挂载点 = agents/paper_orchestra_pipeline.py:1843 的 latex_sanity_check(main_tex) 处 —— 这里 main_tex(已渲染)在 scope,且紧接 if not latex_sanity_report.get("ok"): ... 已是阻断逻辑。
• 把四个检查接入此处:任一命中 → latex_sanity_report["ok"]=False 且 hits 写入 report → 既有阻断生效,禁止出稿。
• 非破坏:不动 require_submission_ready()(证据侧门禁)与现有 latex_sanity_check 既有词表行为;只新增展示侧检查项与 hits 字段。

验收用例(集成)

#	输入	期望输出
M3-1	稿件含 Table ??	latex_sanity_report.ok=False,hits 含 "unresolved reference" + 行号
M3-2	稿件含复读句 ×4	ok=False,hits 含 "boilerplate repetition"
M3-3	稿件干净	ok=True,放行
M3-4(回归)	触发现有 placeholder / FORBIDDEN_LATEX_TERMS 命中	维持原有 ok=False(证明未破坏既有门禁)
M3-5	多处命中(未解析 + 脚手架)	hits 列出全部命中项,不只第一个

DoD

M3-1~5 全绿(挂在 1843),含回归 M3-4;回归基线 tests/test_pipeline_contracts.py、tests/test_vnext_manuscript.py 全绿;收敛分支、未确认不强推。

[Protocol-zero-0]

Completed in PR #29 on branch epic-23-evidence-ledger.

Changed files:

• agents/paper_orchestra_pipeline.py
• tests/test_vnext_manuscript.py

Tests added:

• Render-gate integration for unresolved Table ??.
• Render-gate integration for repeated boilerplate.
• Clean existing bundle path still passes.
• Existing placeholder gate remains blocking.
• Multi-hit case reports both unresolved reference and scaffold placeholder.
• Cross-run identity detection uses rendered main_tex plus state.method_name.

Commands actually run:

• Pre-change baseline: pytest tests/test_pipeline_contracts.py -> 12 passed
• Pre-change baseline: pytest tests/test_vnext_manuscript.py -> 5 passed
• Milestone: pytest tests/test_vnext_manuscript.py -> 10 passed
• Regression: pytest tests/test_pipeline_contracts.py -> 12 passed
• Combined: pytest tests/test_latex_sanity_m2.py tests/test_vnext_manuscript.py -> 18 passed

Non-goals respected:

• Did not modify contracts/pipeline.py or require_submission_ready().
• Did not bypass existing latex_sanity_check.
• Multi-hit behavior is preserved through the report/blockers path.

[Protocol-zero-0]

Done via #29 (squash-merged into master 2026-06-06). 验收用例与回归基线均在 PR #29 全绿。