loopx 0.5.0__tar.gz → 0.5.1__tar.gz

{loopx-0.5.0/loopx.egg-info → loopx-0.5.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: loopx
-Version: 0.5.0
+Version: 0.5.1
 Summary: A lightweight Loop Engineering control plane for long-running agent goals.
 Author: LoopX contributors
 License-Expression: Apache-2.0
@@ -68,8 +68,10 @@ to hand off across turns, tools, and agents.
 
 ## Meet the Personal Agent Workspace
 
-Turn scattered agent sessions into one local-first workspace for Goals,
-attention, conversations, tasks, files, schedules, and recovery.
+Keep long-horizon goals in one local-first workspace. Goals, attention,
+conversations, tasks, files, schedules, and recovery stay durable across days,
+restarts, and harnesses, so work that began last week resumes exactly where
+the last bounded turn stopped instead of living in chat memory.
 
 <a href="docs/assets/personal-workspace/loopx-dashboard-launch.mp4">
   <img src="docs/assets/personal-workspace/loopx-dashboard-tour.webp" alt="Animated tour of the LoopX personal Agent workspace, from one-command launch to Manager overview, Goal task board, protected action preview, and Goal details" width="960">
@@ -100,7 +102,10 @@ npm run dev
 
 [Read the desktop shell guide](apps/desktop/loopx-control-plane/README.md).
 Closing the window stops only the service processes that the shell started;
-existing LoopX services and durable Goal state remain intact.
+existing LoopX services and durable Goal state remain intact. Both entry
+points share the same loopback services and can be started in either order:
+`loopx dashboard` reuses an already-running LoopX Chat service instead of
+starting a second one.
 
 [Watch the full 32-second walkthrough](docs/assets/personal-workspace/loopx-dashboard-launch.mp4)
 · [Read the workspace guide](docs/guides/personal-workspace-user-guide.md)
@@ -314,14 +319,16 @@ LoopX should reuse existing state rather than overwrite it. Keep `.loopx/`,
 | KunlunCode | Run `loopx-kunluncode connect --project . --goal-id <goal-id> --agent-id <registered-agent-id>`, add a bounded todo, then run `loopx-kunluncode run --project .`. | Native Goal Pro through app-server; LoopX writes completion and quota only after strict verification |
 | OpenCode | Install the static command facade; opt in to `--with-goal-bridge` for recurring goals. | OpenCode command facade and explicit goal bridge |
 | Pi | Install the opt-in goal extension with `loopx slash-commands --install --surface pi`, then use `/loopx <task>` from a trusted Pi session. | Visible Pi goal extension gated by LoopX quota (`loopx_goal_activate` + `agent_settled` continuation) |
+| DeepSeek Harness (dsh) | Install `loopx[deepseek-harness]`, prepare a dsh `cordis.yml`, then use the [dsh goal-mode adapter](loopx/dsh_goal_mode/README.md) with `loopx turn run-once`. | Headless dsh segments through LoopX Turn, each tick gated by `quota should-run` |
 | Cursor, shell, or custom runner | Use the installer and `loopx doctor`; connect manually or call LoopX from your runner. | Your shell, scheduler, or runner |
 
 The exact, copy-ready setup messages and host recovery paths live in
 [Getting Started](docs/guides/getting-started.md). Host integrations can inspect
 the [Codex App host command registry contract](docs/reference/protocols/codex-app-host-command-registry-v0.md),
 the [Codex CLI packaged install path](docs/product/runtimes/codex-cli/codex-cli-packaged-install.md),
-the [Claude Code adapter](loopx/claude_goal_mode/README.md), or the
-[KunlunCode native Goal adapter](loopx/kunluncode_goal_mode/README.md).
+the [Claude Code adapter](loopx/claude_goal_mode/README.md), the
+[KunlunCode native Goal adapter](loopx/kunluncode_goal_mode/README.md), or the
+[DeepSeek Harness turn adapter](loopx/dsh_goal_mode/README.md).
 
 For custom runners, start with the
 [minimal custom runtime example](docs/guides/minimal-custom-runtime-example.md)

{loopx-0.5.0 → loopx-0.5.1}/README.md

@@ -39,8 +39,10 @@ to hand off across turns, tools, and agents.
 
 ## Meet the Personal Agent Workspace
 
-Turn scattered agent sessions into one local-first workspace for Goals,
-attention, conversations, tasks, files, schedules, and recovery.
+Keep long-horizon goals in one local-first workspace. Goals, attention,
+conversations, tasks, files, schedules, and recovery stay durable across days,
+restarts, and harnesses, so work that began last week resumes exactly where
+the last bounded turn stopped instead of living in chat memory.
 
 <a href="docs/assets/personal-workspace/loopx-dashboard-launch.mp4">
   <img src="docs/assets/personal-workspace/loopx-dashboard-tour.webp" alt="Animated tour of the LoopX personal Agent workspace, from one-command launch to Manager overview, Goal task board, protected action preview, and Goal details" width="960">
@@ -71,7 +73,10 @@ npm run dev
 
 [Read the desktop shell guide](apps/desktop/loopx-control-plane/README.md).
 Closing the window stops only the service processes that the shell started;
-existing LoopX services and durable Goal state remain intact.
+existing LoopX services and durable Goal state remain intact. Both entry
+points share the same loopback services and can be started in either order:
+`loopx dashboard` reuses an already-running LoopX Chat service instead of
+starting a second one.
 
 [Watch the full 32-second walkthrough](docs/assets/personal-workspace/loopx-dashboard-launch.mp4)
 · [Read the workspace guide](docs/guides/personal-workspace-user-guide.md)
@@ -285,14 +290,16 @@ LoopX should reuse existing state rather than overwrite it. Keep `.loopx/`,
 | KunlunCode | Run `loopx-kunluncode connect --project . --goal-id <goal-id> --agent-id <registered-agent-id>`, add a bounded todo, then run `loopx-kunluncode run --project .`. | Native Goal Pro through app-server; LoopX writes completion and quota only after strict verification |
 | OpenCode | Install the static command facade; opt in to `--with-goal-bridge` for recurring goals. | OpenCode command facade and explicit goal bridge |
 | Pi | Install the opt-in goal extension with `loopx slash-commands --install --surface pi`, then use `/loopx <task>` from a trusted Pi session. | Visible Pi goal extension gated by LoopX quota (`loopx_goal_activate` + `agent_settled` continuation) |
+| DeepSeek Harness (dsh) | Install `loopx[deepseek-harness]`, prepare a dsh `cordis.yml`, then use the [dsh goal-mode adapter](loopx/dsh_goal_mode/README.md) with `loopx turn run-once`. | Headless dsh segments through LoopX Turn, each tick gated by `quota should-run` |
 | Cursor, shell, or custom runner | Use the installer and `loopx doctor`; connect manually or call LoopX from your runner. | Your shell, scheduler, or runner |
 
 The exact, copy-ready setup messages and host recovery paths live in
 [Getting Started](docs/guides/getting-started.md). Host integrations can inspect
 the [Codex App host command registry contract](docs/reference/protocols/codex-app-host-command-registry-v0.md),
 the [Codex CLI packaged install path](docs/product/runtimes/codex-cli/codex-cli-packaged-install.md),
-the [Claude Code adapter](loopx/claude_goal_mode/README.md), or the
-[KunlunCode native Goal adapter](loopx/kunluncode_goal_mode/README.md).
+the [Claude Code adapter](loopx/claude_goal_mode/README.md), the
+[KunlunCode native Goal adapter](loopx/kunluncode_goal_mode/README.md), or the
+[DeepSeek Harness turn adapter](loopx/dsh_goal_mode/README.md).
 
 For custom runners, start with the
 [minimal custom runtime example](docs/guides/minimal-custom-runtime-example.md)

{loopx-0.5.0 → loopx-0.5.1}/loopx/__init__.py

@@ -2,4 +2,4 @@
 
 __all__ = ["__version__"]
 
-__version__ = "0.5.0"
+__version__ = "0.5.1"

{loopx-0.5.0 → loopx-0.5.1}/loopx/agent_onboarding.py

@@ -314,7 +314,9 @@ def _start_instruction(agent_type: str) -> str:
     if agent_type == "deepseek-harness":
         return (
             "Install `loopx[deepseek-harness]`, prepare a dsh cordis.yml, and run "
-            "`loopx turn run-once` with `scripts/dsh_turn_host_adapter.py` as the "
+            "`loopx turn run-once` with the `loopx.dsh_goal_mode` adapter "
+            "(`python -m loopx.dsh_goal_mode`; the legacy "
+            "`scripts/dsh_turn_host_adapter.py` launcher still works) as the "
             "generic-cli host adapter; every tick starts from `quota should-run`."
         )
     if agent_type == "ark-managed-agent":

{loopx-0.5.0 → loopx-0.5.1}/loopx/bootstrap_command_pack.py

@@ -323,7 +323,7 @@ def build_start_goal_host_surface_selection_packet(
         "pi": "Pi LoopX goal extension",
         "gemini-cli": "Gemini CLI driving its own loop through the LoopX skill facade",
         "cursor-agent": "cursor-agent driving its own loop through the LoopX skill facade and MCP server",
-        "deepseek-harness": "DeepSeek Harness automation loop through scripts/dsh_turn_host_adapter.py",
+        "deepseek-harness": "DeepSeek Harness automation loop through loopx.dsh_goal_mode (compat: scripts/dsh_turn_host_adapter.py)",
         "ark-managed-agent": "Ark Managed Agent with one-shot Goal submission",
         "shell": "manual shell or an explicitly configured external scheduler",
         "other-agent": "custom agent host using the returned activation contract",

{loopx-0.5.0 → loopx-0.5.1}/loopx/capabilities/benchmark_toolkit/README.md

@@ -5,6 +5,35 @@ artifact, integrity, and reusable agent-runtime boundaries around benchmark
 experiments. It does not own benchmark-family runners, result ledgers, or scoring
 adapters.
 
+## Source revision admission
+
+A long-running campaign can keep launching from an old installed checkout after
+the tracked branch advances. Pinning the source once at controller startup does
+not prevent that drift. Immediately before each new benchmark admission, obtain
+the current reference head through the runner's provider or network boundary,
+then compare it with the clean local checkout and intended pin:
+
+```bash
+loopx benchmark source-revision-fence \
+  --source-checkout /path/to/pinned-source \
+  --expected-revision "$PINNED_REVISION" \
+  --observed-reference-revision "$OBSERVED_REFERENCE_REVISION" \
+  --require-admitted \
+  --format json
+```
+
+The command succeeds only when all three identities match and the source root has
+no tracked or untracked changes. Its compact receipt records equality, cleanliness,
+and a stable reason code without recording the checkout path or any revision value.
+Invalid input is also reduced to a path-free fail-closed receipt.
+
+The fence is an admission boundary, not a live-run mutation mechanism. A run that
+already passed the fence keeps its immutable revision even if the reference moves
+later; the new head blocks only subsequent admissions until the runner installs and
+pins an updated source. The caller owns the freshness and authority of the observed
+reference value. This capability performs no fetch, provider API call, checkout,
+install, process launch, score write, or submission.
+
 ## Native Codex Goal runtime
 
 Benchmark adapters that use the Codex app-server Goal API should import

{loopx-0.5.0 → loopx-0.5.1}/loopx/capabilities/benchmark_toolkit/__init__.py

@@ -95,6 +95,13 @@ from .run_permissions import (
     compact_run_permission_policy_for_quota,
     validate_run_permission_policy,
 )
+from .source_revision_fence import (
+    BENCHMARK_SOURCE_REVISION_FENCE_SCHEMA_VERSION,
+    BenchmarkSourceRevisionFence,
+    BenchmarkSourceRevisionFenceError,
+    compact_benchmark_source_revision_fence_receipt,
+    inspect_benchmark_source_revision_fence,
+)
 
 __all__ = [
     "BENCHMARK_CANDIDATE_SOURCE_BOUNDARY_SCHEMA_VERSION",
@@ -105,16 +112,20 @@ __all__ = [
     "BENCHMARK_INTEGRITY_POLICY_SCHEMA_VERSION",
     "BENCHMARK_INTEGRITY_QUALIFICATION_SCHEMA_VERSION",
     "BENCHMARK_RUNTIME_INTEGRITY_ATTESTATION_SCHEMA_VERSION",
+    "BENCHMARK_SOURCE_REVISION_FENCE_SCHEMA_VERSION",
     "DEFAULT_RUN_PERMISSION_ALLOWED_ACTIONS",
     "DEFAULT_RUN_PERMISSION_FORBIDDEN_ACTIONS",
     "INTEGRITY_EVIDENCE_CATEGORIES",
     "NATIVE_CODEX_GOAL_PROMPT_SCHEMA_VERSION",
-    "NATIVE_GOAL_LIFECYCLE_SCHEMA_VERSION",
     "NATIVE_CODEX_PROFILE_REQUIRED_SKILL_IDS",
     "NATIVE_CODEX_PROFILE_SCHEMA_VERSION",
+    "NATIVE_GOAL_LIFECYCLE_SCHEMA_VERSION",
+    "PUBLIC_TRAJECTORY_SUMMARY_SCHEMA_VERSION",
     "REQUIRED_RUNTIME_ATTESTATIONS",
     "RUN_PERMISSION_POLICY_SCHEMA_VERSION",
     "RUN_PERMISSION_QUOTA_PROJECTION_SCHEMA_VERSION",
+    "BenchmarkSourceRevisionFence",
+    "BenchmarkSourceRevisionFenceError",
     "DockerContainerBinding",
     "DockerContainerBindingError",
     "NativeCodexGoalPrompt",
@@ -124,15 +135,14 @@ __all__ = [
     "NativeCodexProfile",
     "NativeCodexProfileError",
     "NativeGoalConfig",
-    "NativeGoalLifecycleFacts",
     "NativeGoalEventTransport",
+    "NativeGoalLifecycleFacts",
     "NativeGoalProtocolError",
     "NativeGoalTransport",
     "NativeGoalTurn",
-    "RunPermissionAction",
-    "PUBLIC_TRAJECTORY_SUMMARY_SCHEMA_VERSION",
     "PublicTrajectoryLifecycleState",
     "PublicTrajectorySummaryError",
+    "RunPermissionAction",
     "StdioNativeGoalTransport",
     "attach_native_goal",
     "benchmark_experiment_board_row_key",
@@ -144,6 +154,7 @@ __all__ = [
     "build_run_permission_policy",
     "classify_benchmark_artifact_path",
     "classify_benchmark_candidate_source_path",
+    "compact_benchmark_source_revision_fence_receipt",
     "compact_docker_container_binding_receipt",
     "compact_native_codex_goal_prompt_receipt",
     "compact_native_codex_profile_receipt",
@@ -151,6 +162,7 @@ __all__ = [
     "compact_run_permission_policy_for_quota",
     "default_benchmark_experiment_board_path",
     "filter_public_benchmark_artifact_paths",
+    "inspect_benchmark_source_revision_fence",
     "inspect_native_codex_profile",
     "install_native_codex_profile",
     "materialize_public_benchmark_artifacts",

{loopx-0.5.0 → loopx-0.5.1}/loopx/capabilities/benchmark_toolkit/catalog_entry.py

@@ -24,6 +24,23 @@ BENCHMARK_TOOLKIT_CATALOG_ENTRY: dict[str, Any] = {
     ),
     "entry_command": "loopx benchmark --help",
     "commands": [
+        {
+            "command": (
+                "loopx benchmark source-revision-fence "
+                "--source-checkout <clean-source> "
+                "--expected-revision <pin> "
+                "--observed-reference-revision <observed-head> "
+                "--require-admitted --format json"
+            ),
+            "purpose": (
+                "Fail closed before a new run when the installed source no "
+                "longer matches the caller-observed reference head."
+            ),
+            "write_boundary": (
+                "local Git readback only; caller owns reference observation, "
+                "fetch, installation, and launch"
+            ),
+        },
         {
             "command": (
                 "loopx benchmark experiment-board-show --goal-id <goal-id> "
@@ -90,6 +107,7 @@ BENCHMARK_TOOLKIT_CATALOG_ENTRY: dict[str, Any] = {
         ),
         "required_sequence": [
             "read_experiment_board_before_launch_or_case_selection",
+            "qualify_source_revision_before_each_new_run_admission",
             "upsert_preregistered_or_running_row_when_a_run_starts",
             "upsert_terminal_score_countability_effort_and_insight_status",
             "read_matched_comparisons_before_selecting_the_next_arm",
@@ -181,6 +199,11 @@ BENCHMARK_TOOLKIT_CATALOG_ENTRY: dict[str, Any] = {
         },
     },
     "implemented_protocols": [
+        {
+            "schema_version": "benchmark_source_revision_fence_v0",
+            "module": "loopx.capabilities.benchmark_toolkit.source_revision_fence",
+            "doc": "loopx/capabilities/benchmark_toolkit/README.md",
+        },
         {
             "schema_version": "benchmark_experiment_board_row_v0",
             "module": "loopx.capabilities.benchmark_toolkit.experiment_board",
@@ -220,12 +243,14 @@ BENCHMARK_TOOLKIT_CATALOG_ENTRY: dict[str, Any] = {
     "smokes": [
         (
             "python -m pytest tests/capabilities/test_benchmark_toolkit.py "
-            "tests/capabilities/test_benchmark_experiment_board.py -q"
+            "tests/capabilities/test_benchmark_experiment_board.py "
+            "tests/capabilities/test_benchmark_source_revision_fence.py -q"
         )
     ],
     "docs": ["loopx/capabilities/benchmark_toolkit/README.md"],
     "boundaries": [
         "The toolkit never grants runner, Docker, model, upload, submission, or publication authority; each effect remains separately gated.",
+        "Source revision qualification is read-only and caller-observed: it does not fetch, install, update a checkout, or rewrite an already admitted run.",
         "Raw trajectories are private local inputs to integrity qualification and are never copied into receipts, ledgers, docs, or PR artifacts.",
         "A clean trajectory scan is not isolation proof: runner-owned permission and verifier-order attestations are mandatory and fail closed when absent.",
         "Concurrent Docker runs must bind runtime evidence to one exact job-owned container; image-only discovery is not sufficient.",

loopx-0.5.1/loopx/capabilities/benchmark_toolkit/source_revision_fence.py

@@ -0,0 +1,154 @@
+"""Fail-closed source revision admission for benchmark runs."""
+
+from __future__ import annotations
+
+import re
+import subprocess
+from dataclasses import dataclass
+from pathlib import Path
+from typing import Any
+
+BENCHMARK_SOURCE_REVISION_FENCE_SCHEMA_VERSION = "benchmark_source_revision_fence_v0"
+_FULL_GIT_OBJECT_ID = re.compile(r"(?:[0-9a-f]{40}|[0-9a-f]{64})\Z")
+
+
+class BenchmarkSourceRevisionFenceError(ValueError):
+    """The source checkout or revision observation cannot be trusted."""
+
+
+@dataclass(frozen=True)
+class BenchmarkSourceRevisionFence:
+    """Private exact evidence behind one source admission decision."""
+
+    source_checkout: Path
+    expected_revision: str
+    local_revision: str
+    observed_reference_revision: str
+    source_clean: bool
+    admitted: bool
+    reason_code: str
+
+    @property
+    def local_matches_expected(self) -> bool:
+        return self.local_revision == self.expected_revision
+
+    @property
+    def observed_reference_matches_expected(self) -> bool:
+        return self.observed_reference_revision == self.expected_revision
+
+
+def _normalize_revision(value: str, *, field: str) -> str:
+    revision = str(value).strip().lower()
+    if not _FULL_GIT_OBJECT_ID.fullmatch(revision):
+        raise BenchmarkSourceRevisionFenceError(f"{field}_invalid")
+    return revision
+
+
+def _git_text(source: Path, *args: str) -> str:
+    try:
+        completed = subprocess.run(
+            ["git", "--no-optional-locks", "-C", str(source), *args],
+            check=False,
+            capture_output=True,
+            text=True,
+        )
+    except OSError as exc:
+        raise BenchmarkSourceRevisionFenceError(
+            "source_revision_git_readback_failed"
+        ) from exc
+    if completed.returncode:
+        raise BenchmarkSourceRevisionFenceError("source_revision_git_readback_failed")
+    return completed.stdout.strip()
+
+
+def inspect_benchmark_source_revision_fence(
+    source_checkout: str | Path,
+    *,
+    expected_revision: str,
+    observed_reference_revision: str,
+) -> BenchmarkSourceRevisionFence:
+    """Compare a clean local checkout with a pin and caller-observed ref head.
+
+    The caller owns any network fetch or provider API read used to obtain
+    ``observed_reference_revision``. This helper performs only local readback;
+    it never fetches, checks out, installs, or launches a benchmark run.
+    """
+
+    requested = Path(source_checkout).expanduser()
+    if requested.is_symlink():
+        raise BenchmarkSourceRevisionFenceError("source_checkout_symlink_denied")
+    try:
+        source = requested.resolve(strict=True)
+    except OSError as exc:
+        raise BenchmarkSourceRevisionFenceError("source_checkout_missing") from exc
+    if not source.is_dir():
+        raise BenchmarkSourceRevisionFenceError("source_checkout_not_directory")
+
+    expected = _normalize_revision(expected_revision, field="expected_revision")
+    observed = _normalize_revision(
+        observed_reference_revision,
+        field="observed_reference_revision",
+    )
+    try:
+        repository_root = Path(
+            _git_text(source, "rev-parse", "--show-toplevel")
+        ).resolve(strict=True)
+    except OSError as exc:
+        raise BenchmarkSourceRevisionFenceError(
+            "source_checkout_root_unreadable"
+        ) from exc
+    if repository_root != source:
+        raise BenchmarkSourceRevisionFenceError(
+            "source_checkout_must_be_repository_root"
+        )
+
+    local = _normalize_revision(
+        _git_text(source, "rev-parse", "HEAD"),
+        field="local_revision",
+    )
+    source_clean = not bool(
+        _git_text(source, "status", "--porcelain", "--untracked-files=all")
+    )
+    if not source_clean:
+        admitted = False
+        reason_code = "source_checkout_dirty"
+    elif local != expected:
+        admitted = False
+        reason_code = "local_revision_mismatch"
+    elif observed != expected:
+        admitted = False
+        reason_code = "observed_reference_revision_mismatch"
+    else:
+        admitted = True
+        reason_code = "revision_fence_satisfied"
+
+    return BenchmarkSourceRevisionFence(
+        source_checkout=source,
+        expected_revision=expected,
+        local_revision=local,
+        observed_reference_revision=observed,
+        source_clean=source_clean,
+        admitted=admitted,
+        reason_code=reason_code,
+    )
+
+
+def compact_benchmark_source_revision_fence_receipt(
+    fence: BenchmarkSourceRevisionFence,
+) -> dict[str, Any]:
+    """Return a public-safe decision receipt without paths or revisions."""
+
+    return {
+        "schema_version": BENCHMARK_SOURCE_REVISION_FENCE_SCHEMA_VERSION,
+        "admitted": fence.admitted,
+        "reason_code": fence.reason_code,
+        "source_clean": fence.source_clean,
+        "local_matches_expected": fence.local_matches_expected,
+        "observed_reference_matches_expected": (
+            fence.observed_reference_matches_expected
+        ),
+        "source_path_recorded": False,
+        "revision_values_recorded": False,
+        "network_access_performed": False,
+        "write_performed": False,
+    }

{loopx-0.5.0 → loopx-0.5.1}/loopx/chat_actions.py

@@ -15,14 +15,16 @@ from .chat_action_store import ActionConflictError, ChatActionStore
 from .chat_goal_lifecycle_actions import ChatGoalLifecycleActionMixin
 from .chat_monitor_actions import ChatMonitorActionMixin
 from .chat_store import ChatSessionStore
+from .chat_todo_actions import ChatTodoActionMixin
 from .configure_goal import configure_goal
 from .control_plane.runtime.time import now_utc, parse_timestamp, utc_isoformat
 from .control_plane.scheduler.monitor_todo import monitor_next_due_at
+from .control_plane.todos.contract import require_supported_todo_resume_when
 from .history import load_registry
 from .host_loop_activation import build_host_loop_activation_packet
 from .quota import build_quota_should_run
 from .registry import registry_goals
-from .todos import add_goal_todo, complete_goal_todo, update_goal_todo
+from .todos import add_goal_todo, update_goal_todo
 
 
 CHAT_ACTION_RESPONSE_SCHEMA_VERSION = "loopx_chat_action_response_v1"
@@ -138,7 +140,9 @@ def _monitor_text(parameters: Mapping[str, Any]) -> str | None:
     return None
 
 
-class ChatActionService(ChatGoalLifecycleActionMixin, ChatMonitorActionMixin):
+class ChatActionService(
+    ChatGoalLifecycleActionMixin, ChatMonitorActionMixin, ChatTodoActionMixin
+):
     """Validate previews and route applies through canonical LoopX services."""
 
     def __init__(
@@ -381,9 +385,9 @@ class ChatActionService(ChatGoalLifecycleActionMixin, ChatMonitorActionMixin):
             elif values.get("agent_id"):
                 result["agent_id"] = _opaque(values["agent_id"], field="agent_id")
             if values.get("resume_when"):
-                result["resume_when"] = _text(
-                    values["resume_when"], field="resume_when", limit=240
-                ).lower()
+                result["resume_when"] = require_supported_todo_resume_when(
+                    _text(values["resume_when"], field="resume_when", limit=240)
+                )
             if values.get("successor_todo_ids") is not None:
                 if not isinstance(values["successor_todo_ids"], list):
                     raise ValueError("successor_todo_ids must be a list")
@@ -1183,65 +1187,6 @@ class ChatActionService(ChatGoalLifecycleActionMixin, ChatMonitorActionMixin):
         )
         return {"proposal": stored, "turn": None}
 
-    def _apply_todo_update(
-        self, proposal_id: str, proposal: dict[str, Any], parameters: dict[str, Any]
-    ) -> dict[str, Any]:
-        goal_id = str(parameters["goal_id"])
-        current_fingerprint = self._goal_state_fingerprint(goal_id)
-        if current_fingerprint != proposal.get("expected_state_fingerprint"):
-            stale = self.store.apply(
-                proposal_id, current_state_fingerprint=current_fingerprint, receipt={}
-            )
-            return {"proposal": stale, "turn": None}
-        operation = str(parameters.get("operation") or "edit")
-        if operation == "complete":
-            result = complete_goal_todo(
-                registry_path=self.registry_path,
-                goal_id=goal_id,
-                todo_id=str(parameters["todo_id"]),
-                note=parameters.get("note"),
-                no_followup=bool(parameters.get("no_followup", True)),
-                successor_todo_ids=parameters.get("successor_todo_ids"),
-                agent_id=parameters.get("agent_id"),
-                authority_reason="owner-confirmed typed Chat action",
-                dry_run=False,
-            )
-        else:
-            status = parameters.get("status")
-            if operation == "block":
-                status = "blocked"
-            elif operation == "defer":
-                status = "deferred"
-            result = update_goal_todo(
-                registry_path=self.registry_path,
-                goal_id=goal_id,
-                todo_id=str(parameters["todo_id"]),
-                text=parameters.get("text"),
-                status=status,
-                note=parameters.get("note"),
-                claimed_by=parameters.get("agent_id") if operation == "reassign" else None,
-                resume_when=parameters.get("resume_when"),
-                successor_todo_ids=parameters.get("successor_todo_ids"),
-                agent_id=parameters.get("agent_id"),
-                authority_reason="owner-confirmed typed Chat action",
-                dry_run=False,
-            )
-        todo_id = _opaque(result.get("todo_id"), field="todo_id")
-        receipt = {
-            "receipt_id": _digest({"proposal_id": proposal_id, "todo_id": todo_id})[:32],
-            "outcome": (
-                "todo_completed"
-                if operation == "complete"
-                else "todo_updated" if result.get("changed") else "todo_unchanged"
-            ),
-            "projection_verified": True,
-            "resource_ids": {"goal_id": goal_id, "todo_id": todo_id},
-        }
-        stored = self.store.apply(
-            proposal_id, current_state_fingerprint=current_fingerprint, receipt=receipt
-        )
-        return {"proposal": stored, "turn": None}
-
     def preview(self, request: Mapping[str, Any]) -> dict[str, Any]:
         unknown = set(request) - {
             "action_kind",
@@ -1287,6 +1232,12 @@ class ChatActionService(ChatGoalLifecycleActionMixin, ChatMonitorActionMixin):
             evidence = ["The recoverable Goal and Agent Chat Session is available."]
             permission = "scoped_correction"
         elif action_kind in {"todo.update", "monitor.update"}:
+            if action_kind == "todo.update":
+                canonical_preview = self._run_todo_update(normalized, dry_run=True)
+                if canonical_preview.get("ok") is not True:
+                    raise ValueError(
+                        str(canonical_preview.get("error") or "Todo transition failed canonical dry-run validation")
+                    )
             goal_fingerprint = self._goal_state_fingerprint(normalized["goal_id"])
             if action_kind == "monitor.update" and normalized.get("operation") == "run_now":
                 session_id = normalized.get("session_id")
@@ -1304,7 +1255,11 @@ class ChatActionService(ChatGoalLifecycleActionMixin, ChatMonitorActionMixin):
                 )
             else:
                 fingerprint = goal_fingerprint
-            evidence = ["Canonical LoopX Todo state validated the requested transition."]
+            evidence = [
+                "Canonical LoopX Todo dry-run validated the requested transition."
+                if action_kind == "todo.update"
+                else "Canonical LoopX Todo state validated the requested transition."
+            ]
             permission = "durable_write"
         else:
             fingerprint = self._registry_fingerprint()