More coherent interactive artifacts and review workflows
This release expands Visualize behavior guidance while tightening evidence and release decisions.
- Adds an interactive-state recipe for accessible tabs, forms, dialogs, feedback, focus, and truthful local behavior.
- Gives Artifact Explore and System Explore a shared comparison harness with candidate-scoped controls, frame-aware responsive behavior, and explicit selection boundaries.
- Reworks Animate around frequency, purpose, restraint, interruption, physical continuity, immediate state feedback, and reduced-motion behavior.
- Strengthens Review with explicit coverage states, exhaustive collection before ranking, a report-wide non-blocker cap, context-suppression audit trails, and deterministic grade/verdict precedence.
- Adds evidence tiers and prompt-injection boundaries to Teach so measured, derived, and inferred brand decisions remain distinguishable.
See the Visualize skill, interactive-state recipe, and Explore comparison harness.
Verification and scope
Fresh independent agents exercised identity-authority conflicts, inner-frame responsive previews, accessible interaction behavior, reduced motion, high-frequency commands, finding caps, detector skips, and grade/verdict precedence. The complete local CI-equivalent suite and GitHub CI passed, including Linux validation and Windows Mermaid rendering.
This release publishes the canonical skill plus synchronized npm-skills, Hermes, and Pi/OpenClaw mirrors.