AI analysis grounded in the code graph — computed facts, not vibes · 2026-07-06T05:40:57Z
Taste-skill is a set of portable "Agent Skills" — SKILL.md files installable via the npx skills add CLI (from vercel-labs/agent-skills) — that are meant to bias AI coding assistants (Codex, Cursor, Claude Code) toward better layout, typography, motion and spacing rather than generic AI-generated UI. It also bundles image-generation skills for producing design reference boards (web, mobile, brand kits) intended to be paired with tools like ChatGPT Images before implementation. It targets developers using AI agents for frontend work who want a lightweight, drop-in style guide rather than a full framework.
The star count (57,843 total, +863 in a day) is not explained by the commit history in the graph, which is almost entirely README and sponsor-asset churn (badge contrast fixes, sponsor logo resizing, WebP conversion). The README does reference a "v2 (experimental)" rewrite of the default design-taste-frontend skill and a VS Code marketplace manifest addition (.claude-plugin/), which are the only functionally meaningful changes visible. No release notes were fetched, so the growth is most plausibly attributable to the README's positioning ("Anti-Slop Frontend Framework for AI Agents") and its distribution via the npx skills add ecosystem rather than to code changes evidenced here.
What changed recently, how it's actually built (from the code graph), and whether you should care. Free account — no card, no spam.