visualize v0.4.1
Improves how visualize responds to design feedback and decides when refinement is complete.
- Distinguishes bounded corrections from changed requirements and unresolved structural choices.
- Routes reviews by the underlying problem instead of automatically prescribing simplification for low grades.
- Lets polish finish with few or no edits when rendered evidence supports it.
- Keeps shared brand and token changes outside artifact-refinement authority unless explicitly approved.
- Requires inspected responsive comparisons for Artifact Explore and preserves the approved brief and selection rationale.
- Documents the guiding principles and behavioral validation expectations in CONTRIBUTING.md.
Validated with three independent agents across polish convergence, mobile clipping, token repair, and unresolved reading-order scenarios. All distributed skill mounts are synchronized.