mirror of
https://github.com/Leonxlnx/taste-skill.git
synced 2026-09-19 00:50:24 +08:00
Leonxlnx
4e2cf5afda
Project Architecture: - Moved root SKILL.md into taste-skill/SKILL.md (the original design taste skill) - Created redesign-skill/SKILL.md for upgrading existing projects - Created output-skill/SKILL.md for enforcing complete AI output - Each skill is fully self-contained with no cross-references Redesign Skill (New): - Complete rewrite based on 3 research PDFs (100 AI Design Tells, Creative Web Design Compendium, React UI Component Analysis) - Covers 100+ audit checks across typography, color, layout, interactivity, content, components, iconography, code quality - Includes alignment checks: bottom-aligned CTAs in card groups, consistent vertical rhythm, optical vs mathematical centering - Detects dark/light mode inconsistencies within the same page - Encourages background imagery usage for flat empty sections - Framework-agnostic: works with Tailwind, vanilla CSS, or any styling approach without assuming a specific stack - Upgrade techniques section with motion, typography, layout, and surface improvement patterns - Prioritized fix order from highest-impact to final polish Taste Skill: - Fixed section 7 heading: renamed 'THE 100 AI TELLS' to 'AI TELLS' since the section contains 18 items, not 100 - Removed references/ subfolder (was not relevant to the skill) Research: - Split 2 flat research files into 3 professional subdirectories: root-causes/ (4 files), remediation/ (4 files), findings/ (2 files) - Created research/laziness/ subfolder for topic-based organization - All files rewritten in clean professional English, no emojis - Added research/README.md as navigation index README: - Rewritten in normal language, documents all 3 skills - Added Examples section with split screenshot (floria-top/bottom) - Lists design dial system (variance, motion, density) - Links to research directory Examples: - Added floria-full.webp (original full-page screenshot) - Split into floria-top.webp and floria-bottom.webp for side-by-side display in the README
Research
Background research that informed the skills in this project. Each topic gets its own subfolder.
Topics
LLM Laziness
Why AI models produce incomplete outputs (placeholder code, truncated responses, skipped sections) and documented techniques to prevent it. Covers root causes, parameter fixes, prompt techniques, and experiment data.