Taste Lab
An open-source CLI skill that turns any URL into a design brief plus the reasoning behind each choice, for your AI agent
Taste Lab writes down what a designer gave up, and makes that a required output rather than an optional extra. The clearest evidence is Principle 1 in its own Linear sample: no accent colour, white as the accent against #08090A, with the trade-off stated plainly — you cannot differentiate feature tiers by colour. Token tools like Figma inspect and Stark export values and stop there. If your agent keeps producing UI that is technically correct and tonally wrong, this is the missing file. If you want a native Figma plugin, a browser extension or a no-terminal setup, this is not it — it assumes git, Playwright MCP and a terminal.
Verified 5d ago · liveness 70/100 · cite: rightaichoice.com/tools/taste-lab
- AI agent developers who need design reasoning, not just token values
- Design engineers reverse-engineering a competitor's system rules
- Teams already working in Claude Code, Gemini CLI, Cursor or Windsurf who want taste added to existing rules
- Non-technical users who want a visual design tool or GUI
- Anyone expecting a browser extension for live, real-time design inspection
- Teams needing one consolidated multi-page design system from a single run
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Taste Lab if you need a GUI, a live browser-extension inspector, or a native Figma plugin rather than a terminal skill that writes rule files your agent reads.
The one-time Playwright MCP install downloads Chromium at roughly 100MB, so plan for disk and bandwidth on first setup
Taste Lab is free and open-source, installed by cloning the taste-skill repo — no account or API key is required for the tool itself. The real cost is your existing agent subscription and a one-time ~100MB Chromium download for Playwright MCP. That puts it below paid design-inspection seats, but it is a single-purpose skill rather than a design platform, so budget it alongside your agent tooling, not instead of it.
In short
Taste Lab — An open-source CLI skill that turns any URL into a design brief plus the reasoning behind each choice, for your AI agent. Best for AI agent developers who need design reasoning, not just token values, Design engineers reverse-engineering a competitor's system rules, Teams already working in Claude Code, Gemini CLI, Cursor or Windsurf who want taste added to existing rules. Free to use.
What people actually say about Taste Lab — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
30 mentions across 2 sources (Product Hunt, Lemmy) · researched Jul 3, 2026.
Average across the 1 source that answered — each source counts once, not each post.
- +Reveals design rationale (Taste DNA) not just raw tokens.
- +Generates structured .md and .json output for AI agents.
- +Extracts 20 measurement categories including spacing and shadows.
- +Detects consistent base units and system-level patterns.
- +Automated anti-slop filtering ensures quality output.
- −No motion or interaction capture — static only.
- −CLI-only interface excludes non-developer designers.
- −No multi-site comparison or synthesis feature.
- −No API for integration into existing pipelines.
- −No inconsistency detection for evolving designs.
- • None currently — tool is free and open-source
Viability Score
How well maintained and how widely used is Taste Lab? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Turn any URL into a design context with the /taste slash command
- Four-agent pipeline: Extract Measurements, Detect Patterns, Infer Taste, Observer
- 20 measurement categories cited with exact px / hex / ratio values
- 5–8 system-level design rules per page, each with an Evidence line and Design Goal
- 4 taste principles, each with Trigger, Decision, Reason, Evidence and Trade-off
- At least one Restraint principle documenting a deliberate design omission
- Observer quality gate runs anti-slop grep and JSON validation before writing
- Writes {domain}.md and {domain}.json per run
- Output split into Part 1 Design Map (token set) and Part 2 Taste DNA (4 principles)
- Install as a Claude Code skill at ~/.claude/skills/taste
- Install as a Gemini CLI skill at ~/.gemini/skills/taste
- Writes Cursor rules to .cursor/rules/{domain}-taste.mdc
- Writes Windsurf rules to .windsurf/rules/{domain}-taste.md
- Appends a Design Taste section to CLAUDE.md and writes .github/copilot-instructions.md
- Exports taste-tokens.css and instructions for v0 by Vercel and Figma Make
About Taste Lab
Taste Lab is a free, open-source CLI skill by Sen Lin that turns any webpage URL into a structured design context your AI coding agent can read. Run the /taste slash command on a URL and a four-agent pipeline works the same page through progressively sharper lenses: Extract Measurements pulls 20 categories of objective values (every colour, weight, spacing value, radius and shadow cited as exact px / hex / ratio), Detect Patterns derives 5–8 system-level rules with an Evidence line and a Design Goal, Infer Taste produces 4 principles each carrying Trigger, Decision, Reason, Evidence and Trade-off, and Observer acts as a quality gate running anti-slop grep and JSON validation before writing. At least one of the four principles must be a Restraint — documenting what the designer deliberately left out. Each run writes two files, {domain}.md and {domain}.json: a Design Map of the full token set plus a Taste DNA section. Output rules land where your tool already looks — .cursor/rules/{domain}-taste.mdc for Cursor, .windsurf/rules/{domain}-taste.md for Windsurf, an appended Design Taste section in CLAUDE.md, .github/copilot-instructions.md for GitHub Copilot, .bolt/prompt for Bolt, GEMINI.md for Antigravity, plus taste-tokens.css and instructions for v0 by Vercel and Figma Make, and paste-in text for Lovable's Project Knowledge. Setup is a git clone into ~/.claude/skills/taste or ~/.gemini/skills/taste, a one-time Playwright MCP install that downloads Chromium (~100MB), a restart, then /taste on any URL. It is built for AI agent developers, design engineers and product teams who need agents to make sound design calls on pages they have never seen — not a general design platform, and not a GUI.
Behind the Verdict
The pitch is narrow and the product stays inside it. Taste Lab's distinction is decisions × trade-offs: not "background #08090A, font Inter, radius 6px" but why that near-black instead of pure black, and what was left out. That distinction shows up structurally in the output. Part 1, the Design Map, is the token set — colour system (page background #08090A, Surface-1 rgba(255,255,255,0.02), Surface-2 rgba(255,255,255,0.05), text primary #F7F8F8, text secondary #D0D6E0, panel border #23252A), typography (H1 64px at weight 510 with −1.408px tracking, body 15px at weight 400 with 24px line-height, Inter Variable where 510/590 are variable-only weights), spacing (8px base unit, 72px top section padding, 10px inter-section gap, 12px button horizontal padding), radius (2px tags, 6px the most common primary card value, 8px modal and chat panels, 9999px pill badges) and depth (a 1px inset #23252A ring as the primary depth signal, rgba(0,0,0,0.4) 0 2px 4px reserved for floating cards). Part 2, the Taste DNA, is the part nobody else ships: four principles, each with Trigger, Decision, Reason, Evidence and Trade-off. The pipeline enforces this — Extract Measurements, Detect Patterns, Infer Taste, then Observer, which runs anti-slop grep and JSON validation before anything is written, and Infer Taste is required to produce at least one Restraint principle. The distribution story is also unusually concrete for a side project: rules land in .cursor/rules/{domain}-taste.mdc, .windsurf/rules/{domain}-taste.md, an appended Design Taste section in CLAUDE.md, .github/copilot-instructions.md, .bolt/prompt, GEMINI.md, taste-tokens.css plus instructions for v0 by Vercel and Figma Make, and printed text to paste into Lovable's Project Knowledge. That covers the agent surfaces most teams actually build in rather than one vendor's ecosystem. Where it does not fit: it is a terminal-only skill with no GUI, no browser extension for live inspection, and no native Figma plugin — you get CSS and paste instructions instead. It processes one URL per run, so teams wanting a consolidated multi-page design system will be stitching runs together themselves. Output quality tracks page structure, and single-page apps or heavily dynamic sites are the weak spot since Playwright MCP has to render the page before measurement. The honest framing is the one the product gives itself: a narrow instrument for agent-driven design work, not a general design platform. For that job — getting an agent to reason about a design it has never seen — the trade-off file is the whole product, and it is the file your agent is currently missing.
Researching Taste Lab? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Taste Lab actually fits — and what changes day-one when you adopt it.
You clone the repo into ~/.claude/skills/taste, add Playwright MCP, restart Claude Code, then run /taste https://linear.app before asking the agent to build a new settings page.
Outcome: The agent reads the written {domain}.md and .json, including the Restraint principle explaining why no accent colour exists, so the generated page matches the source site's tonal restraint instead of inventing a purple CTA.
You run /taste against a competitor's marketing site to understand how they hold a near-monochrome look across spacing, radius and depth.
Outcome: You get 20 categories of exact values plus 5–8 system rules with Evidence lines, which you fold into your own .cursor/rules file for the next sprint.
You run /taste per domain and let it write .cursor/rules/{domain}-taste.mdc and .github/copilot-instructions.md alongside the existing instructions.
Outcome: Both agents pick up the taste file on their next run with no pasting, and design decisions stay consistent between the two tools.
Use Cases
- Extract design tokens and the reasoning behind them from a competitor landing page
- Feed a per-domain design brief into an AI coding agent before it generates UI
- Reverse-engineer the system rules and trade-offs of a site whose look you admire
- Study why a minimal near-monochrome interface works, with px/hex evidence for each call
- Build a reference library of restraint patterns by analysing multiple sites
Models Under the Hood
as of 2026-10-14
Limitations
- Taste Lab runs as a command-line skill; there is no graphical interface.
- Setup requires cloning a repo into ~/.claude/skills/taste or ~/.gemini/skills/taste, installing Playwright MCP (a one-time ~100MB Chromium download), and restarting your CLI — beginners without terminal habits will find that steep.
- Output quality depends on page structure and may vary on single-page apps or heavily dynamic sites.
- Each run covers one URL, so producing a multi-page system means running /taste per page and reconciling the results yourself.
- Figma users get CSS plus paste instructions rather than a native plugin.
as of 2026-10-03
Verification history
We have re-verified Taste Lab 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-checked, vendor evidence unchanged
- — re-checked, vendor evidence unchanged
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Taste Lab tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Free
$0
Ideal for
AI agent developers and design engineers who already work in a terminal and want per-domain design reasoning in their agent's rule files
What this tier adds
Starting tier — open-source CLI skill with the full four-agent pipeline, no account or API key required
Where the pricing makes sense
The company stage and team size where Taste Lab's pricing actually pencils out — and where peers do it cheaper.
Taste Lab is free and open-source, installed by cloning the taste-skill repo — no account or API key is required for the tool itself. The real cost is your existing agent subscription and a one-time ~100MB Chromium download for Playwright MCP. That puts it below paid design-inspection seats, but it is a single-purpose skill rather than a design platform, so budget it alongside your agent tooling, not instead of it.
Setup time & first value
How long it actually takes to get something useful out of Taste Lab — broken out by persona, not the marketing-page minute.
Agent developers already comfortable in a terminal: about 10 minutes, dominated by the one-time Playwright MCP install and its ~100MB Chromium download, plus a CLI restart. Design engineers new to skills directories: closer to 20–30 minutes to clone into ~/.claude/skills/taste or ~/.gemini/skills/taste, wire up settings.json for Gemini, and get a first /taste run on a URL. First useful output
Switching to or from Taste Lab
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From manual token extraction: run /taste on the source URL instead of hand-copying hex values into a brief
- →From a Figma inspect token export: add the Trade-off layer Taste Lab writes that the export omits
- →From ad-hoc agent prompts describing a house style: point the agent at the generated {domain}.md and .json rule file instead
- →From per-tool hand-written rule files: let /taste write .cursor/rules/{domain}-taste.mdc, .windsurf/rules/{domain}-taste.md and copilot-instructions.md from one run
- ↗To Figma inspect or Stark: you keep the exported token values but lose the written Trigger/Decision/Reason/Evidence/Trade-off reasoning
- ↗To a general design platform: you gain a GUI and multi-page systems, you give up the agent-readable rule-file output
- ↗To hand-written design guidance: you control every line but lose the anti-slop grep and JSON validation gate
- ↗To a native Figma plugin: you gain in-canvas editing but lose the /taste CLI flow and its per-domain rule files
Integrations
Resources & Guides
Tutorials & Learning
Official links
Tools that pair well with Taste Lab
Common stack mates teams adopt alongside Taste Lab, with the specific reason each pairing earns its keep.
Hue
Open-source skill that turns any brand into a design system Claude Code and Codex can build with
Open Design
Open Design turns the coding agent you already run into a local-first design engine for prototypes, slides and HTML video
Ui Ux Pro Max Skill
Free open-source design-intelligence skill that feeds UI styles, palettes, fonts, and chart rules into Claude Code, Cursor, and other AI coding assistants.
Featured Head-to-Head Comparisons
Taste Lab vs Make Real Tldraw
Choose Taste Lab if you need to extract structured design DNA with reasoning from existing sites for AI agents or design systems. Choose Make Real Tldraw if you want to instantly convert hand-drawn sketches into working prototypes without coding. Both are free but serve opposite ends of the design pipeline: analysis vs. generation.
Taste Lab vs Polycam
Choose Taste Lab if you need to decode a website's design system for AI-driven UI development; it's free and CLI-based. Choose Polycam if you need to create 3D scans, floor plans, or digital twins from real-world objects and spaces; it's a mature freemium app with broad export options.
Taste Lab vs Fusion 360
If you're an AI developer or design engineer who needs to feed design context (colors, typography, spacing, rationale) into LLMs or code agents, Taste Lab is a free, powerful CLI tool. If you're a mechanical or manufacturing engineer creating physical products with integrated CAD/CAM/CAE, Fusion 360 is the industry-standard, paid platform. They solve completely different problems—Taste Lab for digital design extraction, Fusion 360 for physical product development.
Alternatives to Taste Lab
View all
Hue
Open-source skill that turns any brand into a design system Claude Code and Codex can build with
Open Design
Open Design turns the coding agent you already run into a local-first design engine for prototypes, slides and HTML video
Ui Ux Pro Max Skill
Free open-source design-intelligence skill that feeds UI styles, palettes, fonts, and chart rules into Claude Code, Cursor, and other AI coding assistants.
Frequently Asked Questions
Categories
Best-of guides
Used Taste Lab? Help shape our editorial sentiment research.