diff --git a/benchmark/edgebench/run.py b/benchmark/edgebench/run.py index 25e42ced91..7ea2fb4ca2 100644 --- a/benchmark/edgebench/run.py +++ b/benchmark/edgebench/run.py @@ -329,9 +329,7 @@ def main(argv=None): if agent.task_entry == "loopx-planned" else {}), "model": args.model, "effort": args.effort, "timeout_seconds": args.timeout, "loopx_commit": pins[0], "runner_commit": pins[1], - **turn_envelope_receipt(explicit=args.turn_envelope, - mode="heartbeat" if args.worker.startswith("heartbeat-") else "plain", - checkpoint_request=declaration if checkpoint_request is not None else None), + **turn_envelope_receipt(explicit=args.turn_envelope), **({"native_goal_continue": True} if args.native_goal_continue else {}), "task_sha256": hashlib.sha256(task_file.read_bytes()).hexdigest(), "feedback": args.feedback, "internet": task.internet, diff --git a/benchmark/runtime/RUNTIME.md b/benchmark/runtime/RUNTIME.md index 95af5c8129..68a0d89ae8 100644 --- a/benchmark/runtime/RUNTIME.md +++ b/benchmark/runtime/RUNTIME.md @@ -45,6 +45,13 @@ fail before execution. Set it back to `false` for the next run to roll back; do not change a running trial's treatment. Captures contain private Goal context. Transport tests do not establish lower token cost or better model decisions. +Checkpoint capture is independent of this display experiment. A `fresh_replan` +declaration may establish a private capture/handoff socket while +`turn_envelope` remains false: the worker keeps the ordinary dispatcher and quota +input, and only an explicit true setting selects the short view. This replaces +the earlier implicit activation by checkpoint handoff. Do not relabel older +receipts or trajectories as feature-off; qualify their actual input separately. + | Mode | Execution/continuation | LoopX skills and state | | --- | --- | --- | | `plain` | One Codex exec, native Goals disabled | Absent | diff --git a/benchmark/runtime/codex.py b/benchmark/runtime/codex.py index 8af1f0bd14..984ffcc20c 100644 --- a/benchmark/runtime/codex.py +++ b/benchmark/runtime/codex.py @@ -67,19 +67,10 @@ def native_goal(self) -> bool: return self.mode in {"native-goal", "loopx-goal"} -def turn_envelope_receipt(*, explicit: bool, mode: str, - checkpoint_request: dict | None = None) -> dict: - """Project known configuration; a guest path alone proves no activation. - - Callers pass the existing parser's validated declaration. Actual wake - receipts still own whether a handoff socket was successfully established. - These source labels remain local derived receipt values, not admission. - """ +def turn_envelope_receipt(*, explicit: bool) -> dict: + """Record only the explicit display opt-in; checkpoint capture is independent.""" if explicit: return {"turn_envelope": True, "turn_envelope_source": "explicit"} - if mode == "heartbeat" and checkpoint_request is not None and ( - checkpoint_request.get("trigger") == "fresh_replan"): - return {"turn_envelope": True, "turn_envelope_source": "checkpoint_handoff"} return {} diff --git a/benchmark/runtime/harbor.py b/benchmark/runtime/harbor.py index afb72df20f..e61bbed28e 100644 --- a/benchmark/runtime/harbor.py +++ b/benchmark/runtime/harbor.py @@ -145,15 +145,7 @@ def __init__( super().__init__(*args, **kwargs) def _turn_envelope_receipt(self) -> dict: - """Project the effective TurnEnvelope setting, not only the explicit opt-in. - - A checkpoint handoff socket makes the worker capture TurnEnvelope on - fresh_replan execute wakes; per-wake receipts remain authoritative. - """ - return turn_envelope_receipt( - explicit=self.execution.turn_envelope, mode=self.execution.mode, - checkpoint_request=self._checkpoint_declaration, - ) + return turn_envelope_receipt(explicit=self.execution.turn_envelope) def _planning_receipt(self) -> dict: return {"native_goal_continue": True} if self.native_goal_continue else {} diff --git a/benchmark/runtime/worker.py b/benchmark/runtime/worker.py index ea703f02b2..c3f12d2055 100644 --- a/benchmark/runtime/worker.py +++ b/benchmark/runtime/worker.py @@ -67,6 +67,8 @@ def heartbeat_body(env: dict[str, str], turn_id: str, *, native_goal: bool, root = Path(env["LOOPX_WAKE_LOG_DIR"]) / turn_id / "decisions" root.mkdir(mode=0o700, exist_ok=False) command += ["--decision-output-root", str(root)] + if env.get("LOOPX_TURN_ENVELOPE") == "1": + command += ["--turn-envelope"] if decision_handoff_socket is not None: command += ["--decision-handoff-socket", decision_handoff_socket] result = subprocess.run( @@ -417,11 +419,6 @@ def run_once(env: dict[str, str]) -> dict: elif execution.mode in {"heartbeat", "loopx-goal"}: body = heartbeat_body(env, turn_id, native_goal=execution.native_goal, decision_handoff_socket=checkpoint_wake.socket_path) - if checkpoint_wake.socket_path is not None and not execution.turn_envelope: - # The handoff socket enables TurnEnvelope capture without the - # explicit opt-in; the receipt records the effective setting. - receipt["turn_envelope"] = True - receipt["turn_envelope_source"] = "checkpoint_handoff" elif execution.mode == "plain": body = Path(env["LOOPX_TASK_DOC"]).read_text(encoding="utf-8") session_wake = None diff --git a/benchmark/tests/test_native_checkpoint_handoff.py b/benchmark/tests/test_native_checkpoint_handoff.py index 56c3f830b0..c56abcc2af 100644 --- a/benchmark/tests/test_native_checkpoint_handoff.py +++ b/benchmark/tests/test_native_checkpoint_handoff.py @@ -529,8 +529,9 @@ def settle_work(result, expected): result = run_once({**env, "LOOPX_CHECKPOINT_REQUEST": str(request)}) assert result["ok"] and result["checkpoint"]["captured"], result assert "turn_envelope" not in prior and "turn_envelope_source" not in prior - assert result["turn_envelope"] is True - assert result["turn_envelope_source"] == "checkpoint_handoff" + assert "turn_envelope" not in result and "turn_envelope_source" not in result + assert 'Use TurnEnvelope:' not in json.dumps(requests[13]) + assert 'interaction_contract' in guards[0] and 'action_signature' not in guards[0] assert result["checkpoint"]["phase"] == "before_replan" assert result["checkpoint"]["source_turn_id"] == prior["turn_id"] assert result["session"]["session_id"] == prior["session"]["session_id"] @@ -710,6 +711,8 @@ def response(payload, index, message): assert not list(logs.glob("*/decisions")) return assert len(result["checkpoint_handoffs"]) == 1 + assert '--turn-envelope' not in shlex.split(guards[0]) + assert 'turn_envelope' not in result and 'turn_envelope_source' not in result event = result["checkpoint_handoffs"][0] if json.loads(request.read_text())["native_session"] and sys.platform == "darwin": # Unsupported native interruption is refused before any signal; diff --git a/benchmark/tests/test_sforge_runtime.py b/benchmark/tests/test_sforge_runtime.py index cf91fbbe05..4d0854114c 100644 --- a/benchmark/tests/test_sforge_runtime.py +++ b/benchmark/tests/test_sforge_runtime.py @@ -182,11 +182,11 @@ async def installed(self, environment): @pytest.mark.parametrize("explicit,trigger,expected", [ (False, None, None), - (False, "fresh_replan", "checkpoint_handoff"), + (False, "fresh_replan", None), (True, None, "explicit"), (True, "fresh_replan", "explicit"), ]) -def test_sforge_profile_uses_validated_checkpoint_trigger(tmp_path, monkeypatch, explicit, trigger, expected): +def test_sforge_profile_does_not_enable_envelope_for_checkpoint(tmp_path, monkeypatch, explicit, trigger, expected): from types import SimpleNamespace pytest.importorskip("sforge") pytest.importorskip("harbor") diff --git a/benchmark/tests/test_shared_codex_runtime.py b/benchmark/tests/test_shared_codex_runtime.py index ed41f587cd..7f0c5bbaf8 100644 --- a/benchmark/tests/test_shared_codex_runtime.py +++ b/benchmark/tests/test_shared_codex_runtime.py @@ -651,10 +651,10 @@ def test_default_execution_has_no_independent_turn_deadline(tmp_path): (False, None, None), (True, None, "explicit"), (False, "after_wake", None), - (False, "fresh_replan", "checkpoint_handoff"), + (False, "fresh_replan", None), (True, "fresh_replan", "explicit"), ]) -def test_harbor_install_receipt_matches_actual_checkpoint_trigger(tmp_path, monkeypatch, enabled, trigger, expected): +def test_harbor_install_receipt_keeps_capture_independent_from_display(tmp_path, monkeypatch, enabled, trigger, expected): pytest.importorskip("harbor") from benchmark.runtime.harbor import BenchmarkCodex, CodexOffline from benchmark.runtime.checkpoint_wake import CheckpointWake @@ -679,7 +679,7 @@ def test_harbor_install_receipt_matches_actual_checkpoint_trigger(tmp_path, monk # Real declaration parser and socket lifecycle, independent of the # metadata helper. Omitted trigger is legal after-wake capture. assert bool(wake.socket_path) is (trigger == "fresh_replan") - assert bool(enabled or wake.socket_path) is (expected is not None) + assert enabled is (expected is not None) commands = [] command_envs = {} async def no_op(*args, **kwargs): diff --git a/docs/reference/protocols/turn-envelope-v0.md b/docs/reference/protocols/turn-envelope-v0.md index 9708e5bf07..c4b9bf2c4c 100644 --- a/docs/reference/protocols/turn-envelope-v0.md +++ b/docs/reference/protocols/turn-envelope-v0.md @@ -109,7 +109,8 @@ Turn ID. The returned Todo selection command retains the root and requested Other repair/replan guard commands still require the caller to retain those transport flags. A saved decision is not permission to skip revalidation. -The heartbeat renderer can opt into the same transport: +The heartbeat renderer can capture decisions while retaining its ordinary +dispatcher and quota display: ```bash mkdir -m 700 ./private-decisions @@ -118,13 +119,29 @@ loopx --format json heartbeat-prompt --thin --runtime-profile generic_cli \ --decision-output-root ./private-decisions ``` -Execute its guard, read the short action/contract/writeback fields, and use +Capture and display are independent. `--decision-output-root` saves complete +decisions; `--decision-handoff-socket` optionally holds their output for a +caller-owned checkpoint. Neither enables TurnEnvelope. This corrects the old +renderer behavior where requesting capture also selected the short dispatcher. +Existing capture callers that intentionally want the short view must now add +`--turn-envelope` to `heartbeat-prompt`; the generated guard and thin regeneration +command retain that explicit choice. Quota's existing explicit flag is unchanged. + +With that display opt-in, execute its guard, read the short action/contract/writeback fields, and use `detail_ref.full_decision` to read the **same invocation** when selection, replan, capability context or truncation requires detail. Replan history remains with the existing owner; this option neither invents another history store nor chooses a strategy. The renderer rejects full/compact/brief modes, missing Turn identity, native Goal-owned Turns and invalid roots. Stable bootstrap prompts cannot pin a -Turn. Omit the capture option to return to the unchanged full-decision dispatcher. +Turn. Omit `--turn-envelope` to use the ordinary dispatcher, including when +checkpoint capture remains enabled. Captured files retain the full decision in +both display modes. + +中文：捕获和模型输入投影独立。`--decision-output-root` 仅保存完整决策， +`--decision-handoff-socket` 仅保留检查点交接；二者均不隐式开启 TurnEnvelope。 +此前捕获目录同时启用短包的行为已修正。需要短包的调用者须对 +`heartbeat-prompt` 显式传 `--turn-envelope`；省略该开关时仍使用普通输入， +选择、结算、身份与完整私有捕获不变。旧运行的实际输入与设置不得追溯改写。 This is Python filesystem/command transport under the existing quota/context owner, not a new capability or Python decision rule. The shared benchmark diff --git a/loopx/cli_commands/support_control.py b/loopx/cli_commands/support_control.py index c8f7868f6f..56c5b354b6 100644 --- a/loopx/cli_commands/support_control.py +++ b/loopx/cli_commands/support_control.py @@ -400,6 +400,7 @@ def handle_support_control_command( turn_instance_id=args.turn_instance_id, decision_output_root=args.decision_output_root, decision_handoff_socket=args.decision_handoff_socket, + turn_envelope=args.turn_envelope, reward_memory_enabled=reward_memory_enabled, ) if args.bootstrap and payload.get("ok"): diff --git a/loopx/cli_commands/support_control_heartbeat_registration.py b/loopx/cli_commands/support_control_heartbeat_registration.py index d313de0235..0c9e506745 100644 --- a/loopx/cli_commands/support_control_heartbeat_registration.py +++ b/loopx/cli_commands/support_control_heartbeat_registration.py @@ -159,7 +159,13 @@ def register_heartbeat_control_commands( heartbeat_prompt_parser.add_argument( "--decision-output-root", - help="Opt into the short TurnEnvelope with full local captures; requires thin mode and an explicit host-owned Turn ID.", + help="Save full local decisions without changing the display; requires thin mode and an explicit host-owned Turn ID.", + ) + + heartbeat_prompt_parser.add_argument( + "--turn-envelope", + action="store_true", + help="Explicitly use the short TurnEnvelope dispatcher; requires --decision-output-root. Capture alone keeps the ordinary full decision.", ) heartbeat_prompt_parser.add_argument( diff --git a/loopx/control_plane/heartbeat/builder.py b/loopx/control_plane/heartbeat/builder.py index 9e8f35fdba..6fecc8306b 100644 --- a/loopx/control_plane/heartbeat/builder.py +++ b/loopx/control_plane/heartbeat/builder.py @@ -176,6 +176,7 @@ def _heartbeat_prompt_commands( turn_identity_arg: str, decision_output_root: Path | None, decision_handoff_socket: str | None, + turn_envelope: bool, ) -> dict[str, str | None]: quota_guard_command = render_quota_guard_command( goal_id, @@ -279,9 +280,10 @@ def _heartbeat_prompt_commands( capture_arg = " --decision-output-root " + shlex.quote(str(decision_output_root)) if decision_handoff_socket is not None: capture_arg += " --decision-handoff-socket " + shlex.quote(decision_handoff_socket) + display_arg = " --turn-envelope" if turn_envelope else "" for key in ("quota_guard_command", "task_body_quota_guard_command"): - commands[key] = str(commands[key]) + " --turn-envelope" + capture_arg - commands["thin_prompt_command"] = str(commands["thin_prompt_command"]) + capture_arg + commands[key] = str(commands[key]) + display_arg + capture_arg + commands["thin_prompt_command"] = str(commands["thin_prompt_command"]) + display_arg + capture_arg return commands @@ -313,6 +315,7 @@ def build_heartbeat_prompt( reward_memory_enabled: bool = True, decision_output_root: str | Path | None = None, decision_handoff_socket: str | None = None, + turn_envelope: bool = False, ) -> dict[str, Any]: if not (full or compact or brief or thin): thin = True @@ -376,6 +379,7 @@ def build_heartbeat_prompt( native_goal_host=native_goal_host, turn_instance_id=normalized_turn_instance_id, decision_handoff_socket=decision_handoff_socket, + turn_envelope=turn_envelope, ) explicit_agent_scopes = normalize_agent_scopes(agent_scopes) if explicit_agent_scopes: @@ -464,6 +468,7 @@ def build_heartbeat_prompt( turn_identity_arg=turn_identity_arg, decision_output_root=decision_output_root, decision_handoff_socket=decision_handoff_socket, + turn_envelope=turn_envelope, ) cli_preflight = render_cli_preflight(cli_bin=cli_bin) task_body_renderer = _select_task_body_renderer( @@ -499,13 +504,13 @@ def build_heartbeat_prompt( compact_prompt_command=str(commands["compact_prompt_command"]), brief_prompt_command=str(commands["brief_prompt_command"]), thin_prompt_command=str(commands["thin_prompt_command"]), - **({"captured_envelope": True} if decision_output_root is not None else {}), + **({"captured_envelope": True} if turn_envelope else {}), **reward_memory_rule_kwargs, ) task_body = bind_exact_turn_settlement_task_body( task_body, turn_instance_id=normalized_turn_instance_id, - captured_envelope=decision_output_root is not None, + captured_envelope=turn_envelope, ) if fine_grained: task_body = f"{task_body}\n\n{FINE_GRAINED_TURN_RULE}" diff --git a/loopx/control_plane/heartbeat/host.py b/loopx/control_plane/heartbeat/host.py index afcf51cda3..e930e9c4cf 100644 --- a/loopx/control_plane/heartbeat/host.py +++ b/loopx/control_plane/heartbeat/host.py @@ -30,11 +30,14 @@ def resolve_heartbeat_capture_root( native_goal_host: bool, turn_instance_id: str | None, decision_handoff_socket: str | None = None, + turn_envelope: bool = False, ) -> Path | None: """Validate capture transport for an already resolved host-owned Turn.""" if root is None: if decision_handoff_socket is not None: raise ValueError("decision handoff requires an explicit decision capture root") + if turn_envelope: + raise ValueError("TurnEnvelope requires an explicit decision capture root") return None if not thin or full or compact or brief or native_goal_host or not turn_instance_id: raise ValueError("--decision-output-root requires thin mode and an explicit host-owned Turn ID") diff --git a/tests/control_plane/test_heartbeat_envelope_transport.py b/tests/control_plane/test_heartbeat_envelope_transport.py index 03d7a4ef8c..f1af232446 100644 --- a/tests/control_plane/test_heartbeat_envelope_transport.py +++ b/tests/control_plane/test_heartbeat_envelope_transport.py @@ -2,7 +2,10 @@ import json import os import shlex +import socket import sys +import tempfile +import threading from pathlib import Path import pytest @@ -17,15 +20,24 @@ ) -def test_renderer_disabled_parity_and_opt_in_scope(tmp_path): +def test_capture_preserves_full_dispatcher_and_envelope_needs_explicit_opt_in(tmp_path): args = dict(goal_id=GOAL_ID, agent_id=AGENT_ID, runtime_profile="generic_cli", turn_instance_id=TURN_ID, thin=True) assert build_heartbeat_prompt(**args) == build_heartbeat_prompt(**args, decision_output_root=None) - payload = build_heartbeat_prompt(**args, decision_output_root=tmp_path) + ordinary = build_heartbeat_prompt(**args) + captured = build_heartbeat_prompt(**args, decision_output_root=tmp_path) + capture_args = ' --decision-output-root ' + shlex.quote(str(tmp_path)) + assert captured['quota_guard_command'] == ordinary['quota_guard_command'] + capture_args + assert captured['task_body'].replace(capture_args, '') == ordinary['task_body'] + assert '--turn-envelope' not in captured['quota_guard_command'] + with pytest.raises(ValueError, match='capture root'): + build_heartbeat_prompt(**args, turn_envelope=True) + payload = build_heartbeat_prompt(**args, decision_output_root=tmp_path, turn_envelope=True) assert payload['interface_budget']['within_budget'], payload['interface_budget'] assert '--turn-envelope --decision-output-root' in payload['quota_guard_command'] assert 'writeback.settlement_plan.ordered_steps' in payload['task_body'] assert '--decision-output-root' in payload['thin_prompt_command'] + assert '--turn-envelope' in payload['thin_prompt_command'] for overrides in (dict(turn_instance_id=None), dict(thin=False, full=True), dict(runtime_profile='codex_app_ssh_goal')): with pytest.raises(ValueError): @@ -40,8 +52,44 @@ def test_incompatible_worker_rejected(mode): Execution(mode=mode, turn_envelope=True) +@pytest.fixture +def private_handoff(request): + if not request.param: + yield None + return + if os.name != 'posix': + pytest.skip('Requires a private Unix socket') + with tempfile.TemporaryDirectory(prefix='hb-capture-', dir='/tmp') as directory: + endpoint = str(Path(directory).resolve() / 'host.sock') + with socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) as listener: + listener.bind(endpoint) + listener.listen() + listener.settimeout(.1) + stopped = threading.Event() + + def release(): + while not stopped.is_set(): + try: + channel, _ = listener.accept() + except TimeoutError: + continue + with channel, channel.makefile('rb') as stream: + message = json.loads(stream.readline()) + channel.sendall(json.dumps({'released': True, 'sha256': message['sha256']}).encode() + b'\n') + + thread = threading.Thread(target=release, daemon=True) + thread.start() + try: + yield endpoint + finally: + stopped.set() + thread.join(timeout=5) + assert not thread.is_alive() + + @pytest.mark.parametrize('enabled', [False, True]) -def test_real_worker_generated_guard_selection_reentry_and_settle_once(tmp_path, enabled): +@pytest.mark.parametrize('private_handoff', [False, True], indirect=True) +def test_real_worker_generated_guard_selection_reentry_and_settle_once(tmp_path, enabled, private_handoff): project, runtime, registry = _write_fixture(tmp_path) _configure_selectable_alternative(project) logs = tmp_path / 'wake logs' @@ -51,7 +99,7 @@ def test_real_worker_generated_guard_selection_reentry_and_settle_once(tmp_path, LOOPX_REGISTRY=str(registry), LOOPX_RUNTIME_ROOT=str(runtime), LOOPX_PROJECT=str(project), LOOPX_GOAL_ID=GOAL_ID, LOOPX_AGENT_ID=AGENT_ID, LOOPX_WAKE_LOG_DIR=str(logs), LOOPX_TURN_ENVELOPE='1' if enabled else '0') - body = heartbeat_body(env, TURN_ID, native_goal=False) + body = heartbeat_body(env, TURN_ID, native_goal=False, decision_handoff_socket=private_handoff) # Execute the renderer's actual guard, preserving its route and flags. block = body.split('```sh\n', 1)[1].split('```', 1)[0] guard = next(line for line in block.splitlines() if 'quota should-run' in line) @@ -62,18 +110,25 @@ def test_real_worker_generated_guard_selection_reentry_and_settle_once(tmp_path, first = json.loads(result.stdout) assert result.returncode == 0, first root = logs / TURN_ID / 'decisions' - if enabled: + captured = enabled or private_handoff is not None + if captured: captures = list(root.glob('*/decision.json')) assert len(captures) == 1 full = json.loads(captures[0].read_text()) - assert first['action_signature']['matches'] - detail = shlex.split(first['detail_ref']['full_decision']) - assert detail[:2] == ['cat', '--'] # Depends on the saved-detail reader contract. - receipts = _heartbeat_receipt_count(runtime, TURN_ID) - readback = subprocess.run(detail, env=env, cwd=project, capture_output=True, - text=True, check=True) - assert json.loads(readback.stdout) == full - assert _heartbeat_receipt_count(runtime, TURN_ID) == receipts + if enabled: + assert first['action_signature']['matches'] + detail = shlex.split(first['detail_ref']['full_decision']) + assert detail[:2] == ['cat', '--'] # Depends on the saved-detail reader contract. + receipts = _heartbeat_receipt_count(runtime, TURN_ID) + readback = subprocess.run(detail, env=env, cwd=project, capture_output=True, + text=True, check=True) + assert json.loads(readback.stdout) == full + assert _heartbeat_receipt_count(runtime, TURN_ID) == receipts + else: + assert first['interaction_contract'] == full['interaction_contract'] + assert 'action_signature' not in first and 'quota' in first + assert 'Use TurnEnvelope:' not in body + assert '--turn-envelope' not in argv else: assert not root.exists() full = first @@ -82,13 +137,17 @@ def test_real_worker_generated_guard_selection_reentry_and_settle_once(tmp_path, '{todo_id}', ALTERNATIVE_TODO_ID) rc, selected = _run_generated_cli(command, registry_path=registry) assert rc == 0, selected - if enabled: + if captured: paths = list(root.glob('*/decision.json')) assert len(paths) == 2 assert json.loads(captures[0].read_text()) == full selected_full = json.loads(next(p for p in paths if p != captures[0]).read_text()) - assert selected['action_signature']['matches'] - assert selected['writeback']['settlement_plan'] == selected_full['interaction_contract']['cli_channel']['settlement_plan'] + if enabled: + assert selected['action_signature']['matches'] + assert selected['writeback']['settlement_plan'] == selected_full['interaction_contract']['cli_channel']['settlement_plan'] + else: + assert selected['interaction_contract'] == selected_full['interaction_contract'] + assert 'action_signature' not in selected else: selected_full = selected identity = selected_full['heartbeat_receipt']['settlement_identity'] @@ -98,7 +157,7 @@ def test_real_worker_generated_guard_selection_reentry_and_settle_once(tmp_path, rc, replay = _run_generated_cli(command, registry_path=registry) assert rc == 0, replay assert _heartbeat_receipt_count(runtime, TURN_ID) == before - if enabled: + if captured: assert len(list(root.glob('*/decision.json'))) == 3 rc, refresh = _run_cli( registry, runtime, 'refresh-state', '--goal-id', GOAL_ID, '--agent-id', AGENT_ID, @@ -148,7 +207,8 @@ def test_renderer_retains_owned_handoff_without_enabling_it_by_default(tmp_path) for field in ('quota_guard_command', 'thin_prompt_command'): argv = shlex.split(packet[field]) assert argv[argv.index('--decision-handoff-socket') + 1] == endpoint - assert '--decision-handoff-socket when present' in packet['task_body'] + assert '--turn-envelope' not in argv + assert 'Use TurnEnvelope:' not in packet['task_body'] with pytest.raises(ValueError, match='capture root'): build_heartbeat_prompt(**args, decision_handoff_socket=endpoint) with pytest.raises(ValueError, match='invalid decision handoff socket'):