feat(semantics): separate action domains and verify production evidence

.github/workflows/full-public-smokes.yml

@@ -73,6 +73,9 @@ jobs:
       - name: Install smoke runtime dependencies
         run: python -m pip install --disable-pip-version-check "jsonschema>=4.23,<5"
 
+      - name: Install locked TypeScript parser for semantic production checks
+        run: npm ci --ignore-scripts
+
       - name: Preview full-public shard
         run: |
           python3 examples/run-smokes.py \

.gitignore

@@ -37,3 +37,6 @@ goals/**/ACTIVE_GOAL_STATE.md
 goals/**/ACTIVE_GOAL_STATE.md.lock
 /runtime/
 logs/
+
+# Legacy semantic census reports are derived locally, never repository authority.
+/loopx/semantics/inventory_v0.json

docs/development/testing-and-quality.md

@@ -354,12 +354,17 @@ uv run --extra test python -m pytest -q
 uv run --extra test loopx canary premerge --from-git-diff
 # For a fork whose PR base is upstream/main, use this instead:
 uv run --extra test loopx canary premerge --from-git-diff --git-diff-base upstream/main
-# Run one semantic smoke or check its generated inventory:
+# Validate semantics; optionally inspect the full-tree inventory without writing it:
 uv run --extra test loopx canary smoke-suite --script semantic-vocabulary-drift-smoke.py
-uv run python scripts/generate_semantic_inventory.py --check
+uv run python scripts/generate_semantic_inventory.py
 git diff --check
 ```
 
+The semantic inventory is computed from the full tracked tree, not committed.
+Use `--output .local/semantic-inventory.json` only when an exported report is useful;
+`--output <path> --check` checks that explicit report without repairing it.
+词表、owner 与预算继续入库并受检查；结构清单按需计算，无须为普通 PR 补生成文件。
+
 Confirm the interpreter and imported checkout when diagnosing a mismatch:
 
 ```bash

docs/reference/effect-interpreter-packet.md

@@ -26,6 +26,28 @@ decision or turn-settlement logic; they give refactor and test code one
 stable abstraction for reading the effect program shape across packet
 families.
 
+The action slot depends on the packet family:
+
+| Observation | Verdict | `effective_action` |
+| --- | --- | --- |
+| Quota should-run | Quota decision | Existing decision/frontier action string |
+| Turn result | `decision` carries `result_kind` | Always JSON `null` / Python `None` |
+| Journal replay | `replay_legal` or `replay_blocked` | Omitted from the internal replay observation |
+
+Turn-result readers must use `decision` for the result verdict. The result lens
+ignores a host-supplied action; it cannot create a quota decision. The TS result
+type fixes its action to null, while the generic quota type retains its default
+string action. This changes the transient result projection, not persisted
+host results, receipts, or journal plans. See the
+[semantic vocabulary RFC](../architecture/rfcs/semantic-vocabulary-convergence-v0.md#m1-action-domains-and-compatibility)
+for the versioned frontier migration and its compatibility boundary.
+
+动作槽位按 packet family 区分：quota 保留决策/前沿动作字符串；Turn result 的
+判决从 `decision` 读取，`effective_action` 固定为 `null`（Python 为 `None`）；
+journal replay 使用自己的判决，不再输出动作字段。host 提供的 action 不会被
+解释为 quota 决策。这只改变瞬时结果投射，不重写落盘的 host result、receipt
+或 journal plan。
+
 ## Turn Journal Lens
 
 `interpret_turn_journal` reads an existing fenced Turn journal and returns an

docs/reference/glossary.md

@@ -0,0 +1,246 @@
+# Semantic Vocabulary Glossary / 语义词表
+
+<!-- Generated by scripts/generate_semantic_bindings.py; do not edit. -->
+
+This is a generated view of the curated registry, not another authority.
+这是注册表的生成视图；修改定义应编辑 owner 和注册表，然后重新生成。
+
+Equal spellings or value sets do not prove equal meaning. Producers write
+values; consumers interpret or pass them through; only owners define sets.
+同名或相同值集不等于同一语义。生产者写值，消费者解释或透传，owner 定义集合。
+
+See the [RFC](../architecture/rfcs/semantic-vocabulary-convergence-v0.md)
+and its [中文版本](../architecture/rfcs/semantic-vocabulary-convergence-v0.zh-CN.md)
+for scope, compatibility, proof boundaries and migration gates.
+
+## agent_scope_frontier_action
+
+Frontier verdict for an agent-scoped lane. New v1 payloads use agent_scope_frontier.action; the root should-run effective_action projects the same value through its registered disjoint union. Legacy v0 signed payloads remain readable without rewriting.
+
+- Tier / 层级: `kernel`; status / 状态: `canonical`.
+- python: [`AgentScopeFrontierAction`](../../loopx/control_plane/agents/agent_scope_frontier.py).
+- typescript: [`AGENT_SCOPE_FRONTIER_ACTIONS`](../../loopx/control_plane/agents/agent_scope_frontier.generated.ts).
+- Values / 值: `agent_scope_exhausted`, `agent_scope_wait`, `reassignment_required`, `successor_replan_required`.
+
+## delivery_continuity_preemption
+
+Reason a delivery continuity is preempted.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`DELIVERY_CONTINUITY_PREEMPTIONS`](../../loopx/control_plane/turn_driver/delivery_continuity.py).
+- typescript: [`DELIVERY_CONTINUITY_PREEMPTIONS`](../../loopx/control_plane/turn_driver/delivery_continuity.ts).
+- Values / 值: `heartbeat_receipt`, `blocking_work_lane`, `autonomous_replan`, `control_repair`, `delivery_not_allowed`.
+
+## delivery_outcome
+
+Outcome class of a delivered work item.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`DeliveryOutcome`](../../loopx/control_plane/work_items/delivery_outcome.py).
+- typescript: [`DELIVERY_OUTCOMES`](../../loopx/control_plane/work_items/delivery_outcome.ts).
+- Values / 值: `surface_only`, `outcome_gap`, `outcome_progress`, `primary_goal_outcome`.
+
+## delivery_workspace_identity_kind
+
+How a delivery workspace is identified.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`DELIVERY_WORKSPACE_IDENTITY_KINDS`](../../loopx/control_plane/agents/delivery_workspace.py).
+- typescript: [`DELIVERY_WORKSPACE_IDENTITY_KINDS`](../../loopx/control_plane/agents/delivery_workspace.ts).
+- Values / 值: `git_repository`, `local_goal`.
+
+## delivery_workspace_kind
+
+Workspace kind a delivery runs in.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`DELIVERY_WORKSPACE_KINDS`](../../loopx/control_plane/agents/delivery_workspace.py).
+- typescript: [`DELIVERY_WORKSPACE_KINDS`](../../loopx/control_plane/agents/delivery_workspace.ts).
+- Values / 值: `canonical_checkout`, `independent_git_worktree`, `local_goal_workspace`.
+
+## delivery_workspace_requirement
+
+Workspace requirement a settlement asserts before delivery.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`DELIVERY_WORKSPACE_REQUIREMENTS`](../../loopx/control_plane/quota/settlement_workspace_causality.py).
+- typescript: [`DELIVERY_WORKSPACE_REQUIREMENTS`](../../loopx/control_plane/quota/settlement_workspace_causality.ts).
+- Values / 值: `required`, `not_required`, `unknown`.
+
+## effective_action
+
+Compacted should-run verdict carried by status/should-run payloads and the Turn Envelope; the Python enum owns the finite value domain while the wire field remains a string for compatibility.
+
+- Tier / 层级: `kernel`; status / 状态: `merge_candidate`.
+- python: [`EffectiveAction`](../../loopx/control_plane/quota/effective_action.py).
+- typescript: [`EFFECTIVE_ACTIONS`](../../loopx/control_plane/quota/effective_action.generated.ts).
+- Values / 值: `agent_monitor_only`, `agent_workspace_repair`, `automation_prompt_upgrade_required`, `autonomous_replan_required`, `blocked_health`, `blocked_wait`, `boundary_projection_repair`, `capability_bridge_repair`, `control_plane_health_repair`, `control_plane_projection_repair`, `control_plane_repair`, `coordinate_task_bundle`, `external_evidence_observe`, `governed_capability_intent`, `heartbeat_receipt_write_failed`, `heartbeat_settled_skip`, `lark_inbox_reply_due`, `monitor_due`, `monitor_quiet_skip`, `normal_run`, `operator_gate_notify`, `operator_inbox_material_review_due`, `outcome_floor_recovery`, `peer_coordination_blocked`, `quota_skip`, `runtime_user_gate_projection_repair`, `scoped_user_gate_fallback`, `state_projection_gap_repair`, `terminal_no_followup`, `throttled_skip`, `todo_decision_scope_projection_repair`, `unsettled_host_turn_recovery`.
+
+## goal_amendment_class
+
+Class of a proposed Goal amendment.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`GOAL_AMENDMENT_CLASSES`](../../loopx/control_plane/goals/goal_amendment_proposal.py).
+- typescript: [`GOAL_AMENDMENT_CLASSES`](../../loopx/control_plane/goals/goal_amendment_proposal.ts).
+- Values / 值: `lane_route`, `shared_work_graph`, `shared_acceptance`, `protected_authority`.
+
+## goal_amendment_proposal_admission
+
+Admission verdict for a Goal amendment proposal.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`GOAL_AMENDMENT_PROPOSAL_ADMISSIONS`](../../loopx/control_plane/goals/goal_amendment_proposal.py).
+- typescript: [`GOAL_AMENDMENT_PROPOSAL_ADMISSIONS`](../../loopx/control_plane/goals/goal_amendment_proposal.ts).
+- Values / 值: `admitted`, `needs_rebase`.
+
+## goal_amendment_proposal_admission_fact
+
+Fact recorded with a Goal amendment admission verdict.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`GOAL_AMENDMENT_PROPOSAL_ADMISSION_FACTS`](../../loopx/control_plane/goals/goal_amendment_proposal.py).
+- typescript: [`GOAL_AMENDMENT_PROPOSAL_ADMISSION_FACTS`](../../loopx/control_plane/goals/goal_amendment_proposal.ts).
+- Values / 值: `base_state_event_basis_sequence_behind_derived_head`, `base_source_basis_digest_mismatch`, `base_source_basis_unverifiable`, `base_revision_basis_superseded`.
+
+## handoff_mode
+
+Authority handoff mode between agents.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`HandoffMode`](../../loopx/control_plane/coordination/authority_core.py).
+- typescript: [`HANDOFF_MODES`](../../loopx/control_plane/coordination/handoff_mode_policy.ts).
+- Values / 值: `legacy`, `soft_claim`, `hard_lease`.
+
+## lease_action
+
+Authority-core lease mutation verb.
+
+- Tier / 层级: `kernel`; status / 状态: `legacy`.
+- python: [`LeaseAction`](../../loopx/control_plane/coordination/authority_core.py).
+- Values / 值: `acquire`, `renew`, `transfer`, `release`.
+- Compatibility only / 兼容保留: `acquire`, `release`, `renew`, `transfer`.
+
+## loop_disposition
+
+Pure controller verdict for the outer loop after combining the last Turn receipt with the fresh route.
+
+- Tier / 层级: `kernel`; status / 状态: `canonical`.
+- python: [`LoopDisposition`](../../loopx/control_plane/turn_driver/loop_controller.py).
+- Values / 值: `run_now`, `capability_action_required`, `wait`, `stop`, `user_action_required`, `repair`, `replan`, `terminal`.
+
+## receipt_bound_monitor_phase
+
+Receipt-bound phase of a monitor poll settlement.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`ReceiptBoundMonitorPhase`](../../loopx/control_plane/effect_program.py).
+- typescript: [`RECEIPT_BOUND_MONITOR_PHASES`](../../loopx/control_plane/quota/settlement_phase.ts).
+- Values / 值: `poll_due`, `settlement_pending`, `settled`.
+
+## receipt_bound_replay_phase
+
+Receipt-bound phase of a Turn journal replay settlement.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`ReceiptBoundReplayPhase`](../../loopx/control_plane/effect_program.py).
+- typescript: [`RECEIPT_BOUND_REPLAY_PHASES`](../../loopx/control_plane/quota/settlement_phase.ts).
+- Values / 值: `open`, `settlement_pending`, `settled`.
+
+## scheduler_cadence_transition
+
+Legal scheduler cadence state transition.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`SchedulerCadenceTransition`](../../loopx/control_plane/scheduler/state_transition_rules.py).
+- typescript: [`SCHEDULER_CADENCE_TRANSITIONS`](../../loopx/control_plane/scheduler/state_transition_rules.ts).
+- Values / 值: `initial`, `identity_reset`, `retry_unacknowledged_failure`, `hold_active_initial`, `advance_after_interval`, `hold_until_interval`.
+
+## scheduler_host_transition
+
+Legal scheduler host state transition.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`SchedulerHostTransition`](../../loopx/control_plane/scheduler/state_transition_rules.py).
+- typescript: [`SCHEDULER_HOST_TRANSITIONS`](../../loopx/control_plane/scheduler/state_transition_rules.ts).
+- Values / 值: `apply_required`, `host_match_ack_required`, `recorded_failure_suppressed`, `settled`.
+
+## settlement_binding_kind
+
+How a settlement step binds to its receipt.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`SettlementBindingKind`](../../loopx/control_plane/effect_program.py).
+- typescript: [`SETTLEMENT_BINDING_KINDS`](../../loopx/control_plane/effect_program.ts).
+- Values / 值: `todo`, `autonomous_replan`, `unbound`.
+
+## settlement_failure_kind
+
+Typed failure class of a settlement step.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`SettlementFailureKind`](../../loopx/control_plane/effect_program.py).
+- typescript: [`SETTLEMENT_FAILURE_KINDS`](../../loopx/control_plane/effect_program.ts).
+- Values / 值: `invalid_identity`, `receipt_missing`, `identity_mismatch`, `writeback_missing`, `writeback_rejected`, `quota_spend_rejected`, `terminal_closeout_rejected`, `cancelled`, `permission_denied`, `budget_rejected`, `effect_outcome_unknown`.
+
+## settlement_step_kind
+
+Effect-program settlement step executed for one Turn.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`SettlementStepKind`](../../loopx/control_plane/effect_program.py).
+- typescript: [`SETTLEMENT_STEP_KINDS`](../../loopx/control_plane/effect_program.ts).
+- Values / 值: `validation`, `durable_writeback`, `quota_spend`, `terminal_closeout`.
+
+## todo_completion_continuation
+
+Continuation declared by a completing Todo.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`TodoCompletionContinuation`](../../loopx/control_plane/todos/completion_state.py).
+- typescript: [`TODO_COMPLETION_CONTINUATIONS`](../../loopx/control_plane/todos/completion_state.ts).
+- Values / 值: `active_goal`, `successor`, `no_followup`.
+
+## todo_completion_recovery
+
+Recovery path when a Todo completion cannot continue.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`TodoCompletionRecovery`](../../loopx/control_plane/todos/completion_state.py).
+- typescript: [`TODO_COMPLETION_RECOVERIES`](../../loopx/control_plane/todos/completion_state.ts).
+- Values / 值: `same_turn_terminal_closeout`, `lifecycle_reentry_terminal_closeout`.
+
+## todo_decision_scope_granularity
+
+Granularity of a Todo decision scope.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`TODO_DECISION_SCOPE_GRANULARITY_VALUES`](../../loopx/control_plane/todos/contract.py).
+- typescript: [`TODO_DECISION_SCOPE_GRANULARITIES`](../../loopx/control_plane/todos/decision_metadata.ts).
+- Values / 值: `action`, `lane`, `goal`, `project`, `global`.
+
+## todo_decision_scope_kind
+
+Scope kind of a Todo decision.
+
+- Tier / 层级: `cross_runtime`; status / 状态: `canonical`.
+- python: [`TODO_DECISION_SCOPE_KIND_VALUES`](../../loopx/control_plane/todos/contract.py).
+- typescript: [`TODO_DECISION_SCOPE_KINDS`](../../loopx/control_plane/todos/decision_metadata.ts).
+- Values / 值: `private_read`, `write_scope`, `resource`, `production`, `public_claim`, `direction`, `other`.
+
+## turn_result_kind
+
+Typed public outcome of one executed Turn as settled by the Turn Journal.
+
+- Tier / 层级: `kernel`; status / 状态: `canonical`.
+- python: [`LoopXTurnResultKind`](../../loopx/control_plane/turn_driver/transaction.py).
+- typescript: [`TURN_RESULT_KINDS`](../../loopx/control_plane/turn_driver/settlement.ts).
+- Values / 值: `validated_progress`, `validated_completion`, `repair_required`, `replan_required`, `user_action_required`, `wait`, `iteration_failed`, `host_failure`, `validation_failed`, `writeback_failed`, `quota_spend_failed`, `terminal_closeout_failed`.
+
+## turn_route
+
+Typed delivery route derived from a fresh should-run decision before a Host is engaged.
+
+- Tier / 层级: `kernel`; status / 状态: `canonical`.
+- python: [`LoopXTurnRoute`](../../loopx/control_plane/turn_driver/driver.py).
+- Values / 值: `ready_for_host`, `capability_action_required`, `repair_required`, `replan_required`, `user_action_required`, `wait`, `blocked`, `contract_error`.

examples/control_plane/agent-scope-projection-characterization-smoke.py

@@ -258,9 +258,9 @@ def assert_agent_scope_frontier_builder_contract() -> None:
         extra_fields={"cleared_without_successor_handoff_gates": [{"todo_id": "todo_gate"}]},
     )
 
-    assert payload["schema_version"] == "agent_scope_frontier_v0"
+    assert payload["schema_version"] == "agent_scope_frontier_v1"
     assert payload["action"] == "successor_replan_required"
-    assert payload["effective_action"] == payload["action"]
+    assert "effective_action" not in payload
     assert payload["blocks_delivery"] is True
     assert payload["requires_replan"] is True
     assert payload["quiet_noop_allowed"] is False
@@ -302,7 +302,7 @@ def assert_agent_scope_frontier_and_hint_contract() -> None:
     )
 
     assert frontier is not None, frontier
-    assert frontier["schema_version"] == "agent_scope_frontier_v0"
+    assert frontier["schema_version"] == "agent_scope_frontier_v1"
     assert frontier["action"] == "successor_replan_required"
     assert frontier["requires_replan"] is True
     assert frontier["quiet_noop_allowed"] is False

examples/control_plane/work-lane-contract-smoke.py

@@ -11,24 +11,24 @@
 sys.path.insert(0, str(SMOKE_DIR))
 sys.path.insert(0, str(REPO_ROOT))
 
-from loopx.control_plane.scheduler.execution_context import (
+from loopx.control_plane.scheduler.execution_context import (  # noqa: E402 - standalone smoke bootstraps repo imports
     SchedulerRuntimeProfile,
     scheduler_execution_context_for_runtime_profile,
 )
-from loopx.control_plane.todos.contract import (
+from loopx.control_plane.todos.contract import (  # noqa: E402 - standalone smoke bootstraps repo imports
     TODO_TASK_CLASS_ADVANCEMENT,
     TODO_TASK_CLASS_MONITOR,
 )
-from loopx.quota import (
+from loopx.quota import (  # noqa: E402 - standalone smoke bootstraps repo imports
     build_quota_should_run as _build_quota_should_run,
     render_quota_should_run_markdown,
 )
-from loopx.status import (
+from loopx.status import (  # noqa: E402 - standalone smoke bootstraps repo imports
     compact_todo_group,
     compact_post_handoff_run,
     normalize_todo_task_class,
 )
-from work_lane_contract_fixtures import (
+from work_lane_contract_fixtures import (  # noqa: E402 - standalone smoke bootstraps repo imports
     FUTURE_DUE_AT,
     GOAL_ID,
     PAST_DUE_AT,
@@ -1393,8 +1393,9 @@ def assert_peer_requires_reassignment_when_only_other_peer_has_claimed_work() ->
     assert guard["effective_action"] == "reassignment_required", guard
     assert "agent_lane_next_action" not in guard, guard
     frontier = guard["agent_scope_frontier"]
-    assert frontier["schema_version"] == "agent_scope_frontier_v0", frontier
+    assert frontier["schema_version"] == "agent_scope_frontier_v1", frontier
     assert frontier["action"] == "reassignment_required", frontier
+    assert "effective_action" not in frontier, frontier
     assert frontier["agent_id"] == "codex-side-bypass", frontier
     assert "primary_agent" not in frontier, frontier
     assert frontier["candidate_counts"]["current_agent_claimed_advancement_count"] == 0, frontier