What it actually does
Agent-generated interfaces have a recognisable look: everything centred, default type scale, uniform spacing, no motion. Functional and forgettable.
Taste Skill targets exactly that. It is a portable Agent Skill that upgrades layout, typography, motion, and spacing so generated interfaces stop looking like boilerplate. The maintainers call it the anti-slop frontend framework, which is a fair description of the problem.
It also includes image-generation skills for reference boards — web, mobile, and brand kits. The intended workflow is to generate reference frames with an image model, then hand those frames to Codex, Cursor, or Claude Code as visual direction, so the agent is building toward something rather than improvising.
MIT, 75,000 stars, and notably only 53 open issues.
Compared with UI UX Pro Max, which generates a whole design system, this focuses on craft-level quality in the output itself. They address adjacent problems and can complement each other.
Who it's for
- 01
Founders whose product looks like it was built by an agent, because it was
- 02
Engineers who know their frontend work looks generic but cannot articulate why
- 03
Teams competing in crowded markets where visual quality is a differentiator
- 04
Anyone shipping customer-facing interfaces without design support
Where it earns its keep
- Raising the visual quality of agent-generated screens without a designer
- Establishing visual direction with reference boards before building
- Adding considered motion and spacing rather than framework defaults
- Making a landing page look like someone chose the type
- Improving perceived product quality ahead of a fundraise or launch
Use it, or skip it
Reach for it when
- Your interfaces work but look unremarkable
- You have no designer and visual quality is affecting perception
- You want reference-driven building rather than improvisation
- Typography and spacing are where your output visibly falls short
Skip it when
- You have a designer and a design system — this will introduce competing opinions
- Your product is internal tooling where function fully outweighs polish
- You need accessibility guarantees; visual craft and accessibility are not the same thing
- Strict brand guidelines leave no room for the skill's judgement
10 automations
Ideas, not tutorials. Each one is work a team does by hand today.
- 01Marketing
Reference board first
Generate visual reference frames before any code, then build toward them rather than improvising.
- 02Marketing
Landing page quality pass
Run every campaign page through the skill before launch so paid traffic lands on something considered.
- 03Founders
Pre-launch polish
Apply across the product before a launch or fundraise where first impressions carry weight.
- 04Engineering
Motion consistency
Establish consistent transition and animation conventions rather than per-component improvisation.
- 05Marketing
Brand kit generation
Use the image skills to produce a brand kit that then guides every generated interface.
- 06Sales
Demo environment upgrade
Make the demo environment visually strong, since that is what prospects actually judge.
- 07Engineering
Typography standardisation
Fix inconsistent type scales across screens built at different times by different people.
- 08Marketing
Competitive visual audit
Compare your interface against competitors and use the skill to close the most visible gaps.
- 09Engineering
Component refresh
Systematically upgrade your existing component library rather than rebuilding it.
- 10Marketing
Conversion experiment
A/B test an upgraded page against the original to see whether visual quality moves conversion.
Want one of these running by Friday?
LimeDock builds these as real workflows inside your stack — deployed to your cloud, wired into your Slack and CRM, with the code in your repo. You pay a build fee and your own API keys, nothing else.
Source
Repository stats were read from the GitHub API and reflect the last time we refreshed this entry. The editorial breakdown above is LimeDock’s own analysis — we are not affiliated with Leonxlnx.
https://github.com/Leonxlnx/taste-skill