workflow-fix: Step 5a FAMILY_lint omits scripts/pre_split_review_guard.py — half-sync reds the step9c lint pins in pre-#2294 worktrees

Goal

Add scripts/pre_split_review_guard.py to the Step 5a spec-freshness FAMILY_lint membership (the FAMILY_OF map in .claude/skills/issue/steps/09-step-5.md, mirrored in the Step 10d copy if present), so the guard helper and the surfaces that execute/pin it move as one vintage.

The gap

Commit 0ad97f8f8a (issue-2294) updated three coupled surfaces together on main: scripts/workflow_lint.py (new check: the guard CLI must name IMPLEMENTER-MARKER-MISSING), tests/test_workflow_lint_pre_split_guard.py (pin test executing the WORKTREE copy of the helper), and scripts/pre_split_review_guard.py (the helper itself). The first two are FAMILY_lint members (scripts/workflow_lint.py, :(glob)tests/test_workflow_lint*.py); the helper is NOT in any family and NOT covered by the sibling-issue globs (not issue<N>_-named). A worktree cut before 0ad97f8f8a that runs the Step 5a sync therefore gets main's NEW linter + NEW pin test against the FORK-ERA helper — the exact #1824/#1860 half-sync class the family mechanism exists to prevent.

Observed instance (#2305 gate, 2026-08-23)

Worktree issue-2305 (cut ~20 min before 0ad97f8f8a landed) synced the lint family, then its Step 9c gate red TWO nodes green on main: tests/test_workflow_lint.py::test_workflow_lint_default_exits_zero (the no-flags lint exits 1 on the worktree tree: "scripts/pre_split_review_guard.py: no longer names 'IMPLEMENTER-MARKER-MISSING' (#2294)") and tests/test_workflow_lint_pre_split_guard.py::test_review_lens_passes_on_live_tree. Cost: one full 75-min gate round + a manual reconcile (the #2305 session hand-synced the helper with the sync-anchor subject, commit 3d9ce7bc7e on issue-2305).

Fix

1. Add FAMILY_OF["scripts/pre_split_review_guard.py"]="lint" (and the SPECS list entry) in .claude/skills/issue/steps/09-step-5.md; verify test_step10d_family_atomicity_matches_step5a still passes (the Step 10d copy must not drift).
2. Sweep for other same-class strays: any scripts/*.py helper executed or importlib-loaded by a synced pin-test glob but absent from every family (the #1972 selector precedent: select_step9c_tests.py joined FAMILY_lint for exactly this reason).
3. Update the guard (20) exempt/membership lists if the completeness pin requires it.

Provenance

Found by the #2305 orchestrator during its round-1 Step 9c gate (lint probe on the worktree tree + git log origin/main -- scripts/pre_split_review_guard.py). Not a #2305 deliverable — filed separately per the workflow-fix-on-bug protocol.