[docxology]

Prediction

At head bd99405, the check-set should read exactly two rows, both pass:

• gates — pass (ruff check, ruff format, mypy, artifact build + provenance gates, pytest --cov=src --cov-fail-under=90)
• gnn conformance (pinned) — pass (pinned-parser extract + conformance tests)

No other workflow-owned rows are expected; this repository has a single workflow (gates.yml). If GitHub Actions is still billing-locked, both rows will fail in 1-2 s with steps: 0 and the annotation "The job was not started because your account is locked due to a billing issue" — instrument outage, not a verdict on this change.

Change

ruff check is red on main: SIM102 (collapsible nested if) fires twice in _audit_evidence in src/manuscript/claim_support.py — once for the delivery claim type, once for the visual claim type. This collapses the two nested conditionals into their elif guards:

elif row.claim_type == "delivery" and not row.validators:
elif row.claim_type == "visual" and not row.figures:

Each branch appends the same issue message the nested form appended, and no else exists on either inner if, so the fix is behavior-preserving.

Local verification (the repo's own gates, run before opening)

• uvx ruff check src scripts tests — 0 diagnostics (was 1 on main)
• uvx ruff format --check src scripts tests — 280 files already formatted
• uv run mypy src — Success: no issues found in 142 source files
• uv run pytest --cov=src --cov-fail-under=90 -q — 982 passed, coverage 90.99% (floor 90%)
• GNN_ROOT=... uv run python scripts/validate_gnn_spec.py — exit 0, receipt checks 6/6 at the pinned commit 8f293d9b
• tests/test_claim_support.py — 39/39 passed, including test_delivery_row_needs_validators and test_visual_row_missing_figure, which assert the message text and audit outcome of both touched branches

The regenerated output/ artifacts (PNG byte noise from matplotlib version drift) are left uncommitted.