[Zhouyang66639]

Goal And Delivered Outcome

• Outcome basis / optional anchor: the declaration in tests/control_plane/test_compact_identifier_shape_owner.py and the same-shape copies it did not cover; same conversion pattern as the merged periodic-report slug PR refactor(periodic-report): import the public-safe slug shape from its owner #6922.

• Goal/source and gap: the guard declared deletion_service.py as an open opaque site "being restructured in fix(goals): fence deletion against canonical writes #5854"; fix(goals): fence deletion against canonical writes #5854 merged on 2026-10-07, so the declaration went stale. presentation.py:48 and integrity.py:159 still compiled the compact {0,127} body that public_safe_text.py:345 owns, and the compact shape had no tree census (only the opaque body did).

• Observable before → after, with the validation row that proves it: three private compilations become imports of the owner's object; OPEN_OPAQUE_SITES is empty; a new compact census fails on any future restatement. Behavior is unchanged because each call site uses the same compiled object.

• Issue/task and intended base: Related to fix(goals): fence deletion against canonical writes #5854 (merged), refactor(periodic-report): import the public-safe slug shape from its owner #6922 (same conversion pattern), [Bug]: main's Python digest-owner guard fails - goal_source.py and checkpoint_retention.py restate the bare SHA-256 shape #6841 (digest-owner guard); base c97f921e2.

Author Declaration

• Written by: model_agent (Qoder), at the operator's request and review.

Implemented against

• Specification and revision: no written specification; the guard's own declaration, the owner's comment in loopx/public_safe_text.py, and CONTRIBUTING's maintenance-outcome route are the basis.
• Criteria:

Criterion	Disposition	Symbol / path	Test or command
Opaque declaration is cleared after #5854	implemented	loopx/control_plane/goals/deletion_service.py	test_the_opaque_id_body_is_compiled_only_by_the_owner_and_declared_sites
The two compact copies use the owner	implemented	loopx/extensions/presentation.py, loopx/capabilities/benchmark_toolkit/integrity.py	test_each_contract_holds_the_owners_object_and_reports_its_own_field
Compact shape has a tree census	implemented	tests/control_plane/test_compact_identifier_shape_owner.py	test_the_compact_token_body_is_compiled_only_by_the_owner_and_declared_sites

• Self-check before submission: the census finds the compact body only in the owner after the change (three files before), and the opaque body only in the owner (two before). Ran the focused 622-test set, the full tests/architecture suite on base and head (identical failure set), ruff, and git diff --check. I did not run the Windows lane or the full repository suite.

Scope And Continuation

• Completed scope and remaining work: complete within this scope; no behavior change and no successor needed.
• Slice boundary / successor: N/A.

Validation

• Tested revision: 4031b533054e25c17b5c640fc49973d38458d406
• Run state: finished

Check	Command	Result
Guard	python -m pytest -q tests/control_plane/test_compact_identifier_shape_owner.py	17 passed
Focused behavior	python -m pytest -q tests/control_plane/test_goal_activation.py tests/control_plane/test_shadow_native_todo_update_e2e.py tests/control_plane/test_project_registry_m0_contract.py tests/architecture/test_content_digest_single_owner.py tests/presentation tests/capabilities/test_benchmark_study_projection.py tests/capabilities/test_benchmark_behavior_finding.py tests/capabilities/test_benchmark_four_arm_contract.py	622 passed, 8 skipped
Architecture base/head	python -m pytest -q --tb=no tests/architecture on both trees	head 1464 passed / 92 failed; failure set identical to base
Ruff	ruff check <changed files>	pass
Diff hygiene	git diff --check	pass

[loopx-agent]

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

动机

APPROVE — 已核完整提交 4031b53。

维护者修改公共标识语法时，需要分别检查 Goal 删除、扩展展示和 benchmark 完整性三个消费者。此前三个模块仍各自编译相同表达式，修改公共定义后可能留下旧副本，增加后续维护时的遗漏成本。本次让这三个消费者直接使用已有语法对象，并补上重复定义检查。可观察结果是重复定义消失，现有输入、错误和业务处理结果保持一致。本次只收敛语法定义；删除权限、展示隐私规则和 benchmark 资格规则继续由原消费者负责。

改动思路

公共语法已有明确 owner，三个消费者继续拥有各自业务校验，因此直接导入比新建一层验证服务更合适。当前 PR 边界是三处表达式复用、过期例外清理和现有测试中的重复定义普查。

实际调用顺序仍是：删除入口先规范化并检查 Goal id，再检查停止状态和登记关系，最后由原删除服务写入；扩展展示先检查字段和标识，再核对 manifest、内容及 lineage；完整性归约先检查声明语法，再检查 authority、namespace 和 runner 条件。公共表达式只回答字符串形状是否合法，不授予删除、发布或执行权限。选择不修改可以继续运行，但公共语法的下一次调整仍需同步三个副本；只删除测试例外则不能消除这个维护缺口。这里采用已有 owner，未引入新模块、状态、配置或调用步骤。

具体改动

没有独立书面规格。本次判断基于不可变基线 c97f921 的 公共语法定义及其责任说明、已有消费者行为，以及 CONTRIBUTING 的具体维护成果路线。我先读这些基线，再检查补丁；没有把新增测试当作独立规格。

关键代码讲解

• 删除服务的导入和 _require_opaque_id：以 OPAQUE_ID_PATTERN 替代本地编译，保留 trim、字段名和异常。200 字符的语法上限、停止状态门、锁和恢复逻辑均未改动。
• 展示模块的导入 及 _identifier：复用 COMPACT_TOKEN_PATTERN，仍由本模块检查 128 字符、声明的 surface、view schema 和内容。_evidence_reference 的 HTTPS 与敏感字段限制保持独立。
• 完整性模块的导入 及 _validated_task_environment_access：复用同一 compact 对象，原 authorization id、authority、purpose、namespace 等拒绝规则没有移动到通用 owner；其他使用该对象的 protocol/channel/adjudication/runtime 声明也保留原业务限制。
• 现有 owner 测试：清空 opaque 例外，为 compact 增加整树字符串普查，并核对三个消费者持有 owner 对象。将当前两个普查施加于实际基线源码时均拒绝重复定义，当前源码则通过。此检查覆盖指定字面表达式，不是所有等价正则写法的语义检测。

完整差异为 4 个文件、+33/-11，其中生产代码 +6/-5，测试 +27/-6。三个生产改动均为复用现有对象；owner 的表达式和 flags 与基线一致。最新 main 的上述 owner 和三个消费者也未发生额外变化。

对主干的风险

主要风险是把“语法相同”误当作“整个消费者结果相同”。为此我对基线和当前头分别执行了同一组实际 API：每版 21 个边界输入，外加 6 个完整展示投影、7 个完整性归约及 3 个删除场景。比较了完整展示/完整性结果与异常；删除使用隔离的合成登记及真实文件写入，比较业务字段和最终登记、数据状态，保留原始路径、指纹及备份差异。

128/129 与 200/201 字符、空值、前导符号、空白、Unicode 和分隔符没有产生差异。停止 Goal 可删除登记且数据保留；活动 Goal 和穿越输入在写入前拒绝。未声明 surface、禁止的 token 字段、错误 authority、越出 namespace 及缺少 runner 条件也保持原拒绝结果。测试中的 runner 声明是合成输入，仅用于归约语义比较，不证明真实运行隔离或 benchmark 得分资格。

独立本地结果：聚焦组 623 passed、7 skipped；扩展发布/读取生命周期与 benchmark toolkit 组 190 passed。原生 standard premerge 完成 6 项直接检查和 14 项选中 smoke，零失败，但整体返回 manual_review_required，benchmark 敏感路径仍需人工合并审核，不能写成合并门已通过。没有查询或等待 CI，也未重跑作者报告的完整 architecture suite；Windows 和新的安装包未另行验证。首次自建 probe 对未列入禁止词集的 password 字段作了错误预期，已保留失败并改用基线明确禁止的 token 控制，产品源码及标准未改动。

语义与 CI 对齐

该改动复用现有共享语法；两个例外表只属于测试，没有新增公共词表、配置或持久化字段。测试普查执行时会硬性拒绝不符的源码分布，不能称作仅供参考的建议；它也不改变运行时授权。风险选择器、语义词表 smoke 和双语 RFC 差异检查均已实际执行，人工合并要求保持有效。

我的整体评价

这是一项完整、有实际消费者的维护收敛。长程工作路径判断为 preserved：原停止门、登记写入和恢复责任、投影 lineage 及完整性资格边界都保留；没有新增同步状态或反复操作。用户体验判断为 preserved：同一输入返回相同结果或可行动错误，已有请求和权限下新增必要交互为零。可证明的是当前边界的行为保持和重复定义消除；尚未测量后续维护时间或长期净收益。

本次已在相邻 owner 内完成有界简化，无需再增加服务、参数或迁移层。公共语法 owner 与业务策略分工清楚，保留既有导入名称和序列化格式即可；未发现需要阻断本提交的正确性问题。批准适用于该精确头，人工合并审核及仓库合并资格仍独立保留。

English verdict: APPROVE 4031b53. The three consumers reuse the existing identifier-shape owner with unchanged behavior in paired production-API controls. Independent local validation: 623 passed/7 skipped plus 190 passed; 6 direct and 14 selected premerge checks passed. The benchmark-sensitive manual merge hold remains. CI was not consulted; no live benchmark execution or merge was performed.