Verification: the stop gate blocks on a node the work item never entered

The verification node's captured output, one section per row of testing-plan.md. Environment: Python 3.11, uv 0.12.19, uv sync against the committed uv.lock, THE_LOOP_SERVICE_LOCAL=1 for CLI runs.

Verification results

Row	Command	Outcome
T8 (before)	the bugfix's console session on 48cb8e6	check → ok (at phase-selection); check --recompute → UNMET (at design); gate → exit 2, demands design.md
T7	the three modules on the unfixed source (tests kept, source stashed)	13 failed, 58 passed. Every new pointer case fails, the T6 subprocess reproduction among them; the pinned key set fails on the missing pointer
T1, T2	tests/test_harness_gate.py	green; every pre-existing case unchanged
T3–T6	tests/test_gate_pointer_integration.py, tests/test_core_graphs.py	green
T1–T6, T11	the five modules together, final tree	81 passed
T9	cd cli && uv run python -m pytest -q (CI's command)	4684 passed, 1 skipped in 198.77s; working tree clean of test writes afterwards
T10	ruff format --check, ruff check, pyright cli, scripts/validate_config.py, markdownlint-cli2 "**/*.md"	all green: 0 pyright errors, 0 markdown errors, both configs valid
T8 (after)	the same session, fixed source	gate → exit 0; check --recompute → state: … (pointer at phase-selection; position derived from the artifacts); JSON carries "pointer": "phase-selection"

T8: before and after

Before (unfixed):

$ the-loop check issue-1
issue-1: ok (at phase-selection)
$ the-loop check issue-1 --recompute
issue-1: UNMET (at design)
  BLOCK  design
         · required artifact is missing (docs/specs/issue-1/design.md)
$ THE_LOOP_WORK_ITEM=issue-1 python3 hooks/the-loop-gate.py claude; echo "exit=$?"
the-loop: this step is not complete.
  · required artifact is missing (docs/specs/issue-1/design.md)

Fix the above, then finish. (attempt 1/3)
exit=2

After:

$ the-loop check issue-1 --recompute | head -2
issue-1: UNMET (at design)
  state: …/docs/specs/issue-1/work-item-state.json (pointer at phase-selection; position derived from the artifacts)
$ the-loop check issue-1 --recompute --format json | grep -E '"(currentNode|pointer)"'
  "currentNode": "design",
  "pointer": "phase-selection"
$ THE_LOOP_WORK_ITEM=issue-1 python3 hooks/the-loop-gate.py claude; echo "exit=$?"
exit=0

T7: negative control

FAILED tests/test_harness_gate.py::TestThePointerBoundsTheGate::test_a_node_the_item_never_entered_does_not_block
FAILED tests/test_harness_gate.py::TestThePointerBoundsTheGate::test_a_satisfied_walk_up_to_the_pointer_blocks_nothing
FAILED tests/test_harness_gate.py::TestThePointerBoundsTheGate::test_no_recorded_position_is_inconclusive[]
FAILED tests/test_harness_gate.py::TestThePointerBoundsTheGate::test_no_recorded_position_is_inconclusive[None]
FAILED tests/test_harness_gate.py::TestThePointerBoundsTheGate::test_a_pointer_naming_no_node_in_the_report_is_inconclusive
FAILED tests/test_gate_pointer_integration.py::TestTheReportCarriesThePointer::test_recompute_reports_both_positions
FAILED tests/test_gate_pointer_integration.py::TestTheReportCarriesThePointer::test_without_recompute_the_pointer_is_the_current_node
FAILED tests/test_gate_pointer_integration.py::TestTheReportCarriesThePointer::test_no_state_file_is_no_pointer[False]
FAILED tests/test_gate_pointer_integration.py::TestTheReportCarriesThePointer::test_no_state_file_is_no_pointer[True]
FAILED tests/test_gate_pointer_integration.py::TestTheGateOnARealReport::test_a_parked_item_is_not_told_to_write_a_later_phase
FAILED tests/test_gate_pointer_integration.py::TestTheCheckTableSaysBoth::test_a_derived_position_ahead_of_the_pointer_is_named
FAILED tests/test_gate_pointer_integration.py::TestTheStopHookEndToEnd::test_the_ticket_reproduction_lets_the_turn_end
FAILED tests/test_core_graphs.py::test_a_resolving_repo_keeps_exactly_the_keys_it_always_had
13 failed, 58 passed in 1.58s

These cases pass on the unfixed tree too, and they are meant to. They pin behaviour the fix must keep:

• blocking at the pointer;
• a broken node behind the pointer;
• wait before the pointer;
• an unresolved currentNode.

Note: a stale service during verification

The first "after" run went through a control-plane service that had auto-started on the pre-change code. It answered without pointer, and the gate took the R2.3 fallback (the old behaviour) instead of crashing. Stopping the service fixed it. The operator note in reviewer-briefing.md records this.