Four-step analysis pipeline
Processes a URL through four agent roles so the same page is examined from measurement, pattern, taste, and editorial review perspectives.
Taste
Taste is a Claude Code skill that reads a website’s design taste from a URL and turns it into design tokens plus reasoning for AI agents and developer tools.
Overview
Taste is a Claude Code skill for reading the design taste of a website from a URL. It turns a page into design context that an AI agent can use, combining extracted tokens with the reasoning behind the visual decisions.
According to the homepage, each run analyzes a site in four passes: measurements, patterns, taste inference, and an observer review. The result is a compact brief with a complete token set and a set of taste principles explaining what the design chooses and what it gives up.
Features
Design token extraction
Collects exact design values from the page, including colors, spacing, radii, shadows, typography, and other measurements cited in pixels, hex values, and ratios.
Taste reasoning output
Derives 4 taste principles that explain the design logic behind choices, including the trigger, decision, reason, evidence, and trade-off.
Quality check and validation
Runs an anti-slop quality gate that validates JSON before writing the final files.
Dual-file output
Writes two output files for each run: a `.md` brief and a `.json` file named for the analyzed domain.
Use Cases
Reverse engineer a website’s design
Use Taste when you want to inspect an existing site and extract the design tokens and reasoning that appear to shape its visual system.
Feed design context into an AI agent
Use it when you need structured design context that an AI coding agent can turn into implementation decisions or rule files.
Generate reusable design artifacts
Use it to produce a brief and machine-readable JSON file for a specific domain, then reuse that output in later runs or tools.
Standardize analysis across sites
Use it when comparing multiple websites and you want consistent measurement categories and principle-based analysis rather than subjective notes.
Pros and Cons
Pros
- Uses a structured multi-pass workflow instead of a single-pass summary.
- Captures concrete design measurements with exact values rather than approximations.
- Pairs raw tokens with explanation, which can help an AI agent reproduce the design intent.
- Provides outputs in both Markdown and JSON for downstream tool use.
- Includes a quality gate to filter out invalid or low-quality output.
Cons
- The pricing page is unavailable in the collected evidence, so pricing and plan structure are not confirmed.
- The site only explicitly documents setup for Claude Code and Gemini CLI, even though it mentions several other tool targets.
FAQ
What platforms does Taste support?
It appears to be a Claude Code skill that can also be used with Gemini CLI. The homepage also lists output paths for Cursor, Windsurf, GitHub Copilot, Bolt, Antigravity, v0 by Vercel, Figma, and Lovable, but the site only explicitly shows setup examples for Claude Code and Gemini CLI.
What does the output look like?
The homepage describes a four-step workflow: extract measurements from a webpage, detect patterns, infer taste, and run an observer pass that filters the result. The end product is a `.md` file and a `.json` file for the domain that was analyzed.
How do you use it?
The homepage shows a three-step setup for Claude Code or Gemini CLI: clone the skill into the relevant skills directory, install Playwright MCP, then restart the CLI and run `/taste` on a URL.
Who is it for?
The site says the skill can read any website's design taste and produce design context from a URL. It is aimed at AI agents and developers who want structured design tokens and reasoning from an existing site.
Is pricing available?
The pricing page at `/pricing` currently returns a 404, so the site does not provide a published pricing model in the collected evidence.
Quick Facts
- Category
- Developer Tool
- Primary format
- Claude Code skill
- Supported CLI examples
- Claude Code, Gemini CLI
- Source domain
- tastelab.xyz
- Output files
- .md and .json
- Pricing
- Not published in collected evidence
Taste Alternatives
Ghost
Ghost is a terminal-based AI assistant for chatting, code generation, and CLI tasks. Includes free models, supports Linux, macOS, Windows, and is open source.
图像大厨imgcook
imgcook is a design-to-code tool that converts design drafts into front-end code. It supports plugin-based and developer workflows for Sketch, Photoshop, VS Code, and CLI usage.
Ably Chat
Ably Chat is a chat API platform for building custom realtime chat applications. It supports room-based messaging, typing indicators, presence, reactions, and message updates, with usage-based pricing options for different deployment stages.
AakarDev AI
AakarDev AI helps teams manage AI provider access, project-level setups, logs, and analytics from one dashboard. It supports BYOK workflows and lists providers including OpenAI, Google Gemini, Anthropic, Groq, Mistral AI, and Perplexity AI.
DeepMotion
DeepMotion is a web-based AI motion capture and 3D animation platform with Animate 3D for video-to-animation and SayMotion for text-to-animation.
Arduino VENTUNO Q
Arduino VENTUNO Q is an edge AI computer for AI and robotics applications. It combines AI inference and deterministic control on a single board and is designed to work with Arduino App Lab.