[Bug]: Required Python test-shard rejects the shipped Antigravity steward executor #8028

Preflight

• I searched existing issues and discussions for this behavior.
• This report is not an unpatched security vulnerability.
• The report and attachments contain no credentials or private/internal material.

Issue origin

Observed or reproduced in a real environment

LoopX version or commit

a0130633723210a5406dca61ee27aa33fe48b3f3 (main, 2026-11-04)

Host or runtime surface

Not host-specific

LoopX area

Build, packaging, installer, or CI

Problem

The required Python test collection contains a steward executor editor assertion that still expects the pre-Antigravity two-endpoint vocabulary:

assert fields["executor_endpoint"]["options"] == ["codex", "dsh"]

The owning namespace now accepts antigravity in addition to codex and dsh, and the editor deliberately projects sorted(steward_executor_endpoints()). The stale assertion therefore fails even though production configuration behavior is correct. Because test-shard collects the whole Python suite and pytest requires every shard to pass, this false failure blocks the required merge-gate result for Python-changing work and main qualification.

The mismatch was introduced when #7832 added Antigravity as a Steward executor. The related vocabulary refactor in #7746 kept the production owner and editor aligned, but this test retained a second hard-coded vocabulary.

Minimal reproduction

git checkout a0130633723210a5406dca61ee27aa33fe48b3f3
uv sync --extra test
uv run --extra test python -m pytest -q \
  tests/capabilities/test_capability_configuration_ui.py::test_steward_executor_editor_is_machine_only_and_typed

Expected behavior

The test passes because the editor offers exactly the executor ids accepted by the owning steward executor namespace.

Actual behavior

The test fails deterministically:

AssertionError: assert ['antigravity', 'codex', 'dsh'] == ['codex', 'dsh']
At index 0 diff: 'antigravity' != 'codex'
Left contains one more item: 'dsh'
1 failed in 2.86s

Sanitized diagnostics

platform win32 -- Python 3.14.3 -- pytest-9.2.0
rootdir: loopx checkout at a0130633
configfile: pyproject.toml
FAILED tests/capabilities/test_capability_configuration_ui.py::test_steward_executor_editor_is_machine_only_and_typed

The same relevant source files are unchanged between the observed base and current main; the failure is an assertion mismatch, not a host-specific runtime error.

Additional context

• loopx/capabilities/steward_executor/machine_defaults.py accepts antigravity, codex, and dsh.
• loopx/capabilities/configuration_ui.py projects sorted(steward_executor_endpoints()) into the select field.
• tests/test_antigravity_steward_adapter.py independently confirms that antigravity is in the owner vocabulary.
• .github/workflows/python-tests.yml runs the full collection in six test-shard jobs, and the required merge-gate job depends on the aggregate pytest result.

Proposed labels

bug, triage