Audit AI Design Slop
Finds AI design clichés such as stacked decoration, repeated templates, motion for show and fake proof, and ranks what to remove.
$ npx skills add https://craftwork.design/skills/audit-ai-design-slop -a claude-code -y
Installs into .claude/skills/audit-ai-design-slop. Add -g for every project.
SKILL.md
Run a diagnostic, evidence-backed audit. Identify the smallest set of removals or corrections that would improve the interface while preserving its existing direction.
Boundaries
- Do not guess whether AI made the design.
- Do not assign a numeric slop score.
- Do not reject a visual technique in isolation. A gradient, serif, dark theme, glass effect, card, animation, or single-font system can be intentional.
- Do not prescribe a new font, palette, layout, design system, or art direction unless the user explicitly asks for one.
- Do not turn an audit into an implementation task.
- Mark anything outside the visible or provided evidence as unknown.
Audit the Evidence
Inspect the artifact available in the request:
- screenshots and recordings
- rendered pages and relevant viewports
- interaction and state changes
- source code, tokens, assets, and copy
- console or runtime failures when they affect the experience
For every finding, cite a concrete location, component, behavior, or line of copy. Do not report a generic tendency without evidence in the artifact.
