What It Does
taste is a Claude Code skill that answers “is this good?” rather than “does this work?” It evaluates quality across five interconnected dimensions: code consistency, architecture resilience, product composability, design intentionality, and communication clarity.
The skill provides quick verdicts with highest-leverage fixes for individual files, deep dimension-by-dimension analysis for large codebases, taste-informed generation with explicit rationale, and judgment-oriented comparisons between approaches.
The Five Dimensions
Each dimension balances a core tension and asks an ultimate test question:
- Code: Consistency over cleverness — can a tired engineer understand this?
- Architecture: Resilience over elegance — will this survive the next requirement changes?
- Product: Composability over completeness — does removing this make it better?
- Design: Intentionality over decoration — does every pixel earn its place?
- Communication: Teaching over describing — can a newcomer understand intent without asking?
Use Cases
Designers and product managers use taste to review design system components, PRD quality, and product architecture. Engineers leverage it for code and architecture reviews. Teams apply it to compare implementation approaches before committing. It works with code snippets, design files, architecture docs, or product requirements.
Who Benefits
Product designers assessing component intentionality, design system curators evaluating consistency, product managers writing clearer PRDs, and tech leads building resilient systems all benefit from taste’s calibrated judgment framework.