fix(periodic-report): honor canonical stopped Goals

[mikamikasuki]

Goal And Delivered Outcome

• Outcome basis / optional anchor: Reproduced bug report #6076.
• Goal/source and gap: Periodic-report cadence admission and delivery authority checked only the legacy Goal status field, while the canonical activation owner records owner stops in activation.state.
• Observable before -> after: A Goal with status=active and canonical activation.state=stopped invoked the cadence hook and wrote a cadence window; it now registers no hook, writes no window, and resolves no delivery subscription. The regression test failed before the fix and passes on this branch.
• Issue/task and intended base: Closes [Bug]: Periodic reports are admitted for canonically stopped Goals #6076; base main.

Author Declaration

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

Implemented against

• Specification and revision: Goal activation contract, upstream main at 07eb7a1.
• Criteria:

Criterion (spec clause)	Disposition	Symbol / path	Test or command
A stopped Goal projects automatic Turns as paused	implemented	periodic_report_cadence_hooks; _active_delivery_subscription	test_canonically_stopped_goal_does_not_admit_periodic_report_window

• Self-check before submission: Read the canonical activation predicate and writer, periodic-report cadence and delivery paths, activation contract, issue template, and related PR fix(collaboration): honor canonical stopped goals #5742. The diff is limited to these admission guards and one regression test.

Scope And Continuation

• Completed scope and remaining work: Periodic-report cadence admission and standing delivery authority honor canonical Goal stops. Complete within this scope.
• Slice boundary / successor: N/A.

Validation

• Tested revision: 1af1368
• Run state: finished
• Input classes: synthetic, public_fixture

Check kind	Result	Public-safe evidence / limitation
regression_parity	passed	The added stopped-activation regression failed on upstream 07eb7a1 (hook invocation count was 1); it passes on this revision.
unit	passed	uv run pytest -q tests/capabilities/test_periodic_report_pending_intent.py tests/cli_commands/test_periodic_report_intent_capability_chain.py — 45 passed.
static	passed	Ruff passed for the two changed modules and regression test; git diff --check passed.

• Coverage and gaps: The regression verifies no hook invocation, cadence journal mutation, or standing delivery subscription for a canonical stopped Goal. Existing legacy stopped, paused, and archived cases remain covered by the focused periodic-report suite.

Frontend / Visual Evidence

• UI impact: none
• Before: N/A
• After: N/A
• States and viewports shown: N/A
• Source data: none
• Attention review: N/A

Type of Change

• Bug fix
• New feature
• Breaking change
• Refactoring (no functional changes)
• Documentation update
• Test update

LoopX Area

• Control plane (goals, todos, quota, scheduler, registry, runtime)

Technical Direction

• Direction / acceptance reference, when applicable: Shared Goal Authority.

Shared-authority RFC fixture impact

• Production-scale fixture schema: N/A
• Semantic dimensions changed, or reviewed no-impact rationale: Reuses the existing canonical Goal activation predicate; no vocabulary or schema changed.
• Provider conformance arms run: N/A
• Read-only legacy/file/PostgreSQL three-arm rehearsal: N/A

Boundary Checklist

• Neither the diff nor this PR body/comments/attachments disclose private state, credentials, raw traces or verifier output, internal links, or local machine paths.
• I did not duplicate maintainer-owned benchmark work unless a maintainer split out a public issue for it.
• I kept the change scoped to the linked issue/task.
• I completed the visual evidence section for UI changes, or marked UI impact none.
• Every commit includes a DCO Signed-off-by trailer (git commit -s).

[loopx-agent]

Reviewer: model_agent; gpt-6.1-sol; OpenAI; runtime_reported; xhigh

Exact head: 1af1368f2935ab0539dab92340f992f681543b84; immutable base: 07eb7a1288e8a33995493f9d50475f2664fcbf2e. 全部3文件 +43/-3 已完整复核。

动机

已经停止 Goal、希望定期报告也停止的操作者。 停止动作写入正式 activation 状态，但旧 status 仍为 active；旧版仍安排报告窗口，下一轮还要求准备报告。当前补丁使该 Goal 的自动报告立即停止，恢复后继续原窗口。 隔离真实文件后端和公共 quota CLI 证实：两种正式 stopped 表达都返回 paused，报告 hook、窗口和 pending intent 不再产生；注册后停止及已有窗口停止也不再推进，恢复保留原窗口。 范围是既有定期报告的停止准入和待办解析；没有新增能力开关、前端操作、消息发送或安装升级，不宣称修复所有 Goal 消费者。

独立验收依据 [Bug]: Periodic reports are admitted for canonically stopped Goals，spec_revision issue-6076-body-sha256:a1164122c136622b14779b070fe38ad1a842285a3e8f5006e15dcc7eb858fef2。条款标识（原 issue 无编号，按其原文分句命名）：no_automatic_report；no_hook_window_intent；no_standing_authority。逐条实现与当前源码/真实调用对应；不从测试数、作者完成声明或旧评审继承结论。 正式 activation 的契约另以 07eb7a1288e8a33995493f9d50475f2664fcbf2e 的 docs/book/en/chapters/core-state-machines.md 停止自动 Turn 条款核对；已合并 #5742 明确把 periodic-report 留给独立消费者后续。

改动思路

停止约束是已有 Goal activation 正确性规则，三个消费者复用同一 owner；只在初始化检查会遗漏注册后的停止，因此执行前重读与待报告解析都必须保留。 本 PR 只补既有 periodic-report 的三个停止检查和一个回归测试；合并前还须更新三个已移动的 registry I/O 清单行号。

复用现有 owner，既有能力设置与停止入口没有新页面/确认/重填步骤；CLI 的观察仍消费真实状态。当前 private rehearsal 只用合成隔离 registry/runtime，不修改在线 Goal，不执行付费模型/消息发送。对相邻边界做未来维护检查：不为三处检查再新建 helper/枚举；现有 typed owner 就是最小一致性来源。

具体改动

cadence_runtime.py 在构造 hook 时先读正式 activation，在 journal 准入回调里再次读取当前 Goal，防止 hook 注册后才发生 stop。pending_intent.py::_active_delivery_subscription 同样消费既有 goal_is_stopped，使已安排窗口在停止后也不能继续自动生成。旧 stopped/paused/archived status 兼容检查仍保留。test_periodic_report_pending_intent.py 加26行测试，覆盖正式 stopped 下 dispatch=0、journal=None、subscription=None，没有引入新状态、CLI、写入权或第二状态 owner。

正路径：owner activation -> quota should-run -> cadence registration/admission -> pending report; stop -> no new report; resume -> original pending window。负路径：construct hook -> change canonical authority to stopped -> dispatch -> resolve_current_subscription -> no journal mutation/no pending; explicit resume -> same frozen window。同 registry 的另一个活动 Goal 与停止后新建的活动 Goal 仍独立获得窗口；并未把一 Goal 的停止扩为全局禁用。正式 activation/state 缺失按旧 active 兼容、坏版本仍拒绝，不靠文案判断。

对主干的风险

当前5组 focused tests 101 passed。独立同输入 base/head 12场景：活动正例，正式/投影停止，legacy三态，能力关闭，非注册 reporter，坏 activation，同 registry 既有/新 Goal，两种停止时序；head全部通过，base暴露6个场景的不变量失败。公共 quota should-run 对 active/两种 stopped/关闭均 exit0；停止时 paused/skip、无 periodic-report hook/intent/journal，保留的 manager-context 空 hook 不等于报告被触发。停止已有窗口后 consume 返回 no_pending_intent，恢复不换 window_id、不重写窗口。能力关闭时三个生产消费者的完整返回值与 base 相同；正例及公共决策条款只归一化隔离 fixture 路径。

当前必需 premerge 的5项选择中4项通过，direct5项通过；semantic 一项实际失败。准备 root TypeScript 依赖后，失败已收敛为三个 committed registry I/O 位置：cadence 初始读27→28、执行回调52→57、pending authority232→233。290 site 的身份没有新增；immutable base 同一个完整 semantic 检查通过，head明确报三处 metadata changed。因此不是依赖缺失造成的剩余误报。[P2] 运行默认 uv run python scripts/generate_project_registry_io_manifest.py，保留既有分类并提交三处行号更新，再跑 semantic-vocabulary-drift-smoke 和原生 canary。 当前 regression101 passed 不替代这个仓库要求。

首次 reviewer 用了没有入口的 python -m loopx，修正到项目实际 loopx.entrypoint 后重跑全部对照；本批报告 fixture 原先少 domain/adapter 导致全局健康失败，补合法通用 fixture 字段并显式 generic_cli 后公共调用成功。额外外部路径 manifest 输出探针写出文件后因相对路径打印失败，不能当成清单合格；当前结论来自基线/头同一个完整检查与 committed 元数据三处差异。

按 changed risk 执行足够有界反例；没有请求每个平台/付费交付/长期 soak。已读召回的恢复/RFC经验，但这里有原始 issue 和真实消费者，所以不套用旧 verdict，也不把“读到经验”记为因果学习收益。未读取、轮询或等待 GitHub CI。

我的整体评价

当前一个 P2 阻塞：必须同步清单。停止行为本身已用真实调用证明修复；不把基线可工作的必需检查变红归为环境噪声。 长程方向在本切片是正向：停止后不再重复安排报告、消耗注意力；恢复还能接回原窗口，减少遗漏/重复。效率收益是避免本来不该发生的工作，没有测量吞吐、跨领域多 agent 持续收益。 范围是既有定期报告的停止准入和待办解析；没有新增能力开关、前端操作、消息发送或安装升级，不宣称修复所有 Goal 消费者。 批准/修复意见与安装、合并资格独立；本轮未合并或升级。

English verdict: REQUEST_CHANGES - 1af1368. Canonical stop is now honored at registration, fresh admission and pending authority;101 tests and12 independent source scenarios pass, with stop/resume and scoped/off controls. The required full semantic check still fails for3 stale registry-I/O line metadata entries while the immutable base passes. Regenerate the existing classified manifest and rerun semantic/premerge; runtime correctness alone cannot waive this gate.