fix(replan): exclude monitors from agent long-chain thresholds

[Jim-jimu]

Goal And Delivered Outcome

• Goal/source and gap: An Agent with 9 claimed advancement Todos and 19 continuous monitors receives a long_todo_chain obligation because the old alternative threshold counts all 28 open Todos as backlog.
• Observable before → after: New Agent-lane long-chain obligations require at least 15 claimed open advancement Todos. The 9+19 and 14+6 mixed workloads stay runnable without this obligation; 15 advancement Todos still trigger it. The regression and real-entrypoint checks below prove both boundaries.
• Issue/task and intended base: Closes Long-chain replan treats continuous monitors as research backlog #5001. Base: main at d64c4d3772400c958d51b773f1e1e165cbaca801.

Scope And Continuation

• Complete within this scope: update the existing TypeScript frontier owner, focused regressions, and the protocol, state-machine and bilingual migration documentation. This intentionally retires the 20-claimed-open trigger for new Agent-lane observations.
• Shared unclaimed candidates remain selectable without imposing an owned-lane duty. Unscoped Goal thresholds, monitor due/no-change rules, source completeness, and exact ACK/rearm behavior retain their contracts. Historical open-count checkpoints keep predecessor recovery.
• The related simplification removes the obsolete scoped open-count observation variant and selection calculation. No second rule owner or new abstraction is introduced.
• Slice boundary / successor: N/A; this completes the reported threshold correction. Maintainer review and merge are required.

Validation

• Tested revision: 934e704065c3c66f0d773334ee12766b971710c5
• Run state: finished
• Input classes: synthetic, public_fixture
• Environment: Python 3.13.12 and Node.js 22.22.3.

Check kind	Result	Public-safe evidence / limitation
regression_parity	passed	With production code at 90f21a5299188d54f984a5313e774c9ac48d6595, the same updated tests fail in two TS cases and the 9+19 / 14+6 managed-CLI cases because a long-chain obligation is emitted. All pass on the tested revision.
unit	passed	node --no-warnings --experimental-strip-types --test tests/control_plane_ts/frontier_revision.test.ts tests/control_plane_ts/replan_semantics.test.ts: 24 passed. Covers mixed workloads, ownership isolation, monitor-growth ACK stability, complete-source fences, and historical checkpoint recovery.
integration	passed	After merging upstream #4980, test_python_runtime.test.ts, monitor_successor.test.ts and quota_monitor_poll_commit.test.ts: 29 passed. Confirms the updated Python selector and adjacent monitor tests work with the contribution.
real_entrypoint	passed	93 focused Python tests across frontier rules/scope, canonical FileAuthorityStore, successor causality, monitor routing and CLI projections. Includes tests/control_plane/test_long_chain_projected_closeout.py: real CLI admission and replay of all three mixed workloads, plus the existing durable writeback/spend/rearm journey. All state is disposable.
static	passed	npm run typecheck:control-plane, Ruff on the changed Python test, and git diff --check.
integration	passed	python examples/control_plane/cli-output-budget-regression-smoke.py.
integration	passed	Standard loopx canary premerge --from-git-diff --git-diff-base origin/main: gate passed; 19 selected checks and 4 direct checks executed, with zero blocking failures and one advisory failure disclosed below. Public-boundary scan passed. Here origin/main is the upstream repository.
static	failed	The advisory maintainability check reports the existing module budget finding in loopx/extensions/lark/goal_topic_runtime.py. Reproduced on unmodified baseline 90f21a529; no budget was changed.
static	failed	Configured mypy reports no-any-return at loopx/control_plane/coordination/local_authority_shadow_projection.py:100. The identical error reproduces on unmodified baseline 90f21a529 with mypy 1.20.2.

• Coverage and gaps: Final focused suites, CLI budget and premerge were rerun after incorporating the latest upstream base. The two baseline findings affect unchanged files. Full repository, Windows, PostgreSQL and live-model qualification were not run; this change does not alter provider transactions or storage.

Frontend / Visual Evidence

• UI impact: none. CLI and managed-Turn projections consume the shared threshold decision; no setting, editor or frontend component changes.
• Before: N/A.
• After: N/A.
• States and viewports shown: N/A.
• Source data: none for visual evidence.
• Attention review: N/A; no visual surface changed.

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)
• Benchmark boundary (adapters, runners, verifiers, scoring, evidence)
• Capability or extension (providers, adapters, skills)
• Public docs or presentation surface (README, protocols, dashboard)
• Build, packaging, installer, or CI
• Host or runtime integration

Technical Direction

• Core control-plane hardening: correction within the existing typed frontier owner. Migration checkpoints describe the new threshold; no broader RFC completion is claimed.

Shared-authority RFC fixture impact

• Production-scale fixture schema: unchanged.
• Semantic dimensions changed: Agent-lane long-chain admission excludes monitors. Focused mixed-workload cases cover the new boundary; existing complex File-store frontier/ACK cases remain green. No stored record, provider protocol, capacity or retention assumption changes.
• Provider conformance arms run: existing real FileAuthorityStore frontier/ACK regressions; no full multi-provider conformance run.
• Read-only legacy/file/PostgreSQL three-arm rehearsal: N/A; no promotion, runtime routing or compatibility projection change.

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 (including .loopx/, .codex/goals/, and live ACTIVE_GOAL_STATE.md).
• 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).

[huangruiteng]

动机

#5001 描述的混合负载（9 项已认领推进任务、19 项持续监控）会被旧的“20 项已认领 open Todo”分支误判为长链，强制打断本可继续的 Agent lane。相同真实 CLI 回归在基线 d64c4d3 失败，在本 head 通过；15 项推进任务仍触发，因而这次改动完成了该 issue 的 Agent-lane 目标。

改动思路

沿用现有 TypeScript frontier 决策 owner：Python 只投影持久 Todo 的 task_class、claim 与 frontier 计数，evaluateLongTodoChain 决定阈值，goal-frontier/配额消费 obligation。不新增状态或第二套规则。Agent lane 只看自己已认领的推进任务；无 Agent 的 Goal 总览仍保留原 selectable-open 口径。历史 20-open checkpoint 只在读取/后继恢复路径保留。

具体改动

关键代码讲解

• frontier_revision.ts:234 evaluateLongTodoChain：把 20-open 分支限定在 agent === null，Agent lane 的新触发只剩 15 项 claimed advancement；同时收窄新 observation 的 count_kind。
• frontier_revision.ts:149 successorCheckpoints：保留历史 open-count 前置义务恢复，仍要求唯一新后继、完整匹配源和由 Python 消费方核对的 origin obligation ID。
• TypeScript 测试覆盖 9+19、14+6、15+19、监控增长后的 ACK 稳定性及历史 checkpoint；真实 CLI 测试覆盖首读和 replay。协议、状态机、迁移 RFC 中英两版同步披露旧/新默认值。

对主干的风险

未发现阻塞项。最重要的误回归是监控增长再次让 15 项以下的 Agent lane 进入强制 replan；定向 TS 24/24、真实 CLI/Python 51/51、typecheck 和标准 premerge 均通过。premerge 的 19 项检查无阻塞失败，另有一项未改动 Lark 文件的既有 maintainability advisory，应继续记录但不归因于本 PR。未跑全仓、Windows、PostgreSQL 或 live-model；本变更不修改 provider 写入/存储。

语义与 CI 对齐

这次是已披露的默认行为变更，复用现有 long_todo_chain vocabulary、TypeScript 决策 owner 和持久 ACK 格式；无新权限或可选功能的 default-off 语义。Goal 策略 wait_for_ci=false，本评审未将远端 CI 作为证据或等待条件。

我的整体评价

APPROVE（exact head 934e704）。修复比在 Python 再做一次监控减法更局部，生产改动可逆，历史读侧兼容有针对性测试。剩余风险是无 Agent 总览有意保留 20-selectable-open 阈值，以及未做持久化 pre-upgrade 活跃义务的现场 replay；它们不阻塞 #5001 的 Agent-lane 修复。相邻边界的前瞻性简化已体现在收窄新 observation union，未看到需要并入的更大重构。

English verdict: APPROVE - exact head 934e704; no blocking finding; 24 TS and 51 Python tests, typecheck, and risk-based premerge passed (one inherited advisory).