Summary

• make generate_threejs_factory.py use the orchestrator's authoritative review-completion gate
• prevent manually edited screenshot-only review history from unlocking later sculpt passes
• keep valid, fully reviewed passes unlockable and update the lock guidance to mention the required visual comparison evidence

Problem

The generator had a weaker local copy of review_completes_pass. A manually edited review entry with action=continue and only visualEvidence.renderScreenshot could unlock the next pass, even though sculpt_pass_orchestrator.py correctly rejected the same entry because it lacked a comparison image, an accepted AI vision score, and passing critical feature reviews.

Verification

• Added a regression test and confirmed it failed before the fix because the generator returned exit code 0.
• Confirmed the regression test passes after delegating to the orchestrator gate.
• Confirmed a valid review created through append_sculpt_review.py still unlocks structural-pass generation.
• python3 -m unittest discover -s scripts/tests -v — 20 tests passed.
• ruff check scripts/generate_threejs_factory.py scripts/tests/test_pipeline.py — passed.
• python3 -m compileall -q scripts — passed.
• git diff --check — passed.