[hhyykk]

Goal And Delivered Outcome

Three canonical Todo recovery cases failed with state_file_missing: their setup deleted the entire registry-declared Goal file while trying to remove legacy Todo display. The shipped Goal health guard correctly refused the input, so these cases never qualified the intended resume behavior.

A shared fixture now preserves the same Goal definition and removes only Todo display after canonical capture. The original monitor-repair, generation-wait and capacity-recovery assertions remain. The matrix includes legacy Markdown, File and SQLite. A separate File/SQLite check proves that missing Goal source still denies work, canonical Todos remain readable, and restoring Goal prose restores healthy readback without recreating Todo display.

• Basis: the requested architecture study against the TypeScript migration RFC; independently reproduced qualification defect.
• Intended base: main at 973914e8cb713ee176737c95750aedb3072b9ea8.
• Scope: one test module and paired RFC clarification. Production code, behavior, permissions and defaults are unchanged.

Author Declaration

• Written by: model_agent, OpenAI Codex (GPT-6 family).

Implemented against

• Specification: TypeScript migration RFC section 6 and the existing missing declared Goal source guard.

Criterion	Disposition	Evidence
Qualify actual caller semantics	implemented	Real public quota should-run subprocesses reach resume/capacity checks after Todo display removal.
Preserve source/admission failures	implemented	Canonical data does not waive missing Goal source; both providers retain refusal and restore behavior.
Avoid parallel semantic owners	implemented	Only fixture construction is shared; production owners and rules are unchanged.
Full Goal-source cutover / runtime payoff	out_of_scope	This repairs bounded qualification. No runtime improvement or completed migration claim.

Scope And Continuation

Complete within this fixture correction. _quota_goal owns source construction, canonical capture and Todo-display removal for three scenario families. Its callers keep their independent behavioral assertions. This makes source ownership local to one fixture interface with real File/SQLite adapters.

The migration economics are production LOC 0, bridge LOC 0, new RPCs 0. Duplicated fixture setup is retired; this does not claim product-code deletion or another domain migration.

48 open PR file inventories were checked. No other PR modifies test_resume_planning_projection.py. Shared RFC edits are in the correctness section, separate from the caller/scope payoff checkpoints in #6989/#6990/#6992/#6993/#6995 and the canonical readback work in #6874. No dependent product change is bundled.

Validation

• Tested revision: 6db21f5e3839cc8a8cf715de7c8d86404326d482; test source is identical to the tested working tree.
• State: finished. Inputs: synthetic/public fixtures only. Linux, Python 3.13, qualified Node 24.

Check kind	Result	Evidence / limit
regression_parity	passed	Original main's resume module: 18 passed, 3 failed with missing Goal source. Corrected module plus related health/coordination checks: 53 passed.
real_entrypoint	passed	Public CLI subprocesses from the checkout's editable installation exercise monitor repair, generation waits, capability changes and missing-source/restore. Goal prose bytes remain unchanged by reads.
real_backend	passed	Actual File/SQLite stores use the existing canonical fixture factory; no provider mocks.
regression_parity	passed	Removing canonical initialization in a private copy of the final test module makes all 6 canonical recovery cases fail while all 3 legacy cases pass. This rules out qualification from the retained Goal prose alone.
static	passed	Ruff, diff checks and paired RFC gate; test-guard and docs-guard review completed.
manual	not_run	Wheel/sdist were not rebuilt; production and packaging sources are unchanged. This is source-checkout qualification, not installed release qualification.

Reproduce:

uv sync --extra test
uv run --extra test pytest -q tests/control_plane/test_resume_planning_projection.py tests/control_plane/test_contract_health.py tests/control_plane/test_canonical_goal_channel_coordination.py
uv run --extra test ruff check tests/control_plane/test_resume_planning_projection.py
python3 scripts/check_rfc_bilingual.py --base origin/main --head HEAD

The missing-Goal scenario remains as explicit refusal/recovery coverage; the original resume assertions are retained. File-only promoted variants become explicit File/SQLite variants; the legacy variants remain. Runtime transport and privacy-audit candidates had no sufficient demonstrated payoff and were left unchanged. No performance, model-backed outcome, production promotion, or full asynchronous lifecycle qualification is claimed. CI and maintainer review remain separate.

Frontend / Visual Evidence

UI impact: none. No product interface or screen changes. Before/after, viewport, preview and craft review: N/A. Input data: synthetic/public fixtures.

Shared-authority RFC Fixture Impact

• Shared fixture tests/fixtures/control_plane/coordination_production_scale_v0.json, schema loopx_coordination_production_scale_fixture_v0: unchanged.
• Added qualification dimensions: SQLite parity for these imported Todo recovery cases and explicit Goal-source absence/restoration. No provider-neutral field, canonical operation, capacity or retention rule changes.
• Actual local provider arms: File and SQLite. The fixture factory seeds canonical state; this does not exercise production promotion itself.
• PostgreSQL and read-only legacy/File/PostgreSQL rehearsal: not run; no provider, routing, promotion or compatibility-projection implementation changes.

Type Of Change / Area

Test fixture refactoring, integration coverage and bilingual RFC clarification. Area: control-plane qualification. The runtime source model and health guard are retained.

Boundary Checklist

• Public-safe fixtures and aggregate evidence; no private source data, raw logs, credentials, internal links or local machine paths.
• Open PR scope checked; no maintainer-owned benchmark work duplicated.
• All original scenario assertions retained; source ownership is explicit.
• Both RFC languages changed and the bilingual diff gate passes.
• UI impact marked none.
• Every commit carries a DCO sign-off.

[loopx-agent]

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

APPROVE — 修正了真实的夹具来源错误，保留原恢复断言和健康门禁；已有 147/145 架构预算失败作为独立集成阻塞保留。

动机

维护 canonical Todo 恢复行为及配额准入测试的开发者。
旧夹具为了移除旧 Todo 展示，删掉了仍被注册表声明的整个 Goal 文件，导致三个恢复用例先被来源健康检查拒绝，根本没有走到要验证的恢复分支；新夹具保留同一 Goal 正文，只移除已捕获的 Todo 展示。
现在真实 File、SQLite 和旧 Markdown 调用能分别验证 monitor 修复、generation 等待与 capacity 恢复，同时另有用例确认 Goal 文件真的缺失时仍拒绝执行。
本 PR 修正测试与双语规范说明，不改产品默认、来源健康门禁或权限，也不宣称完整 Goal 迁移、安装发行资格或长期收益。

改动思路

Keep Goal source ownership in one fixture and preserve real production health/resume owners. Source prose and canonical Todo authority are distinct; no fake alternative parser, rule owner or runtime bypass is introduced.
Complete fixture correction and bilingual explanation; bounded recovery qualification only. Full Goal-source migration, packaged promotion, model adoption and complete-cost/performance acceptance remain with existing owners.
这里有两个不同来源：Goal 正文仍由原声明文件提供，Todo 已由 canonical store 提供。修复把前者保留，把后者的旧展示移除；没有用多余正文复制 Todo 来替代真实后端，也没有绕过源文件缺失检查。共享部分只是夹具构造，业务期望仍各自写在原用例里。

具体改动

评审依据在读 diff 前固定为基线 typescript-control-plane-migration-v0.md §6 及原有 test_contract_health.py 的来源拒绝规则。C6-caller-semantic-oracle、C6-missing-source-refusal、C6-canonical-provider-parity 均 implemented；C6-production-promotion 明确 out_of_scope。新增中英文段落语义一致：capture 后只移除 Todo 展示，Goal 真缺失仍拒绝，完整来源迁移另行验收。

关键代码讲解

1. _quota_goal 用既有 factory 初始化真实 File/SQLite canonical authority，再将同一文件写回原 Goal 正文；旧 Markdown 分支保留原 Todo 区段。它只统一重复 setup。
2. test_public_quota_distinguishes_monitor_repair_from_generation_wait 保留原判断：等待常驻 monitor 完成需要修复依赖；等待其 generation 变化在相同 generation 时继续 quiet。capacity 用例也保留 compiler 未提供/已提供时的两次真实 CLI 检查。
3. test_canonical_todos_do_not_replace_a_missing_declared_goal_source 刻意删除 Goal 来源，核 state_file_missing、不允许执行且 canonical Todo 仍有 1 条，再恢复完全相同正文，确认健康恢复且不重建 Todo 展示。

对主干的风险

此类测试修复的主要风险是“变绿但失去 oracle”。独立复现原基线模块是 18 过、3 项来源缺失失败；新头相关恢复、健康、协作用例 53 项通过。同一个未修改的新夹具分别对不可变基线和新头产品执行，各 11 项通过、保留 16 份完整真实 CLI 观察。两种 canonical provider 的缺失来源都退出 1 并拒绝 normal delivery，恢复后仍读原有 Todo；所有读取前后 Goal 文件字节一致。另在私有副本只去掉 canonical 初始化，六个 canonical 恢复变体全失败，三个旧格式正例仍过；相同私有执行环境里的原样控制则全过，说明没有借保留正文冒充 canonical 数据。

标准 premerge 的 4 项直接检查和 15 项选中检查、编译、Ruff、双语 RFC 与公开边界检查通过。不过独立补跑根架构门禁，基线和新头都为同一 test_top_level_module_count_stays_at_the_pinned_budget 失败、5 项通过：147 对 145。两份完整失败签名一致，预算测试、pin 和整个产品模块集合逐字节未改；这是已有集成阻塞，未算作通过，也未要求本夹具 PR 修改预算。#7000 是该失败的独立修复提案。未查 CI、未重建发行包；这里不声称产品安装或完整迁移资格。

我的整体评价

本头 goal_achieved 仅指完整夹具纠正和相应规范说明。long_horizon=preserved，user_experience=improved：维护者能够用正确来源验证真实恢复，原健康拒绝和权限保留。不是从测试数推导模型收益、性能提升或迁移 payoff。全仓库既有健康、恢复和 canonical factory 已复用；同作者当前 10 个 open PR 的完整文件清单只有本 PR 改这个测试模块，其余共享 RFC 的运行时改动不重复这些断言。局部 setup 合并有维护价值，未见一次性 scaffold 或重复 smoke 批量制造；没有贡献限制升级依据。没有新产品状态或并行 owner，现有架构预算红灯和其它 PR 的原 hold 均保留。

English verdict: APPROVE — head 6db21f5e3839cc8a8cf715de7c8d86404326d482 repairs canonical recovery fixtures while preserving the declared Goal health guard and original resume assertions.53 focused tests and standard local premerge pass; identical corrected fixtures qualify real File/SQLite CLI behavior on base/head, with a discriminating6-fail/3-pass canonical-capture mutation. The unchanged147/145 architecture failure remains a separate integration hold. No merge or installed-release claim.