scanned May 24, 2026

Impeccable

impeccable.style

ai-mlVisit site →

Impeccable teaches AI deep design knowledge and provides commands to steer design results.

60/100

Tier 4 · Agent-Ready

Content answers62/100
Protocol plumbing56/1009 of 16 checks pass
Re-audit2585h ago· 1 credit

Scored by asking 15 questions a buyer of a ai-ml product asks, then grading this site’s own pages: answered, hedged (partial or vague), or silent (no page answers it). How scoring works

15 buyer questions · graded against the site’s own pages

6 of 15 come back answered. 5 get a hedge. 4 meet silence.

6 answerable5 half-answers4 silent

Goes silent · limits · high importance

We're migrating a massive legacy UI with thousands of variables. What's the upper limit on how many design tokens the Distill command can process in a single run before it errors out or truncates?

Goes silent · limits · high importance

Overdrive seems like it could blow through my API budget fast. Is there a built-in cap on how many design variations it will generate per request, or can I set that threshold myself?

This report is public. Own impeccable.style? Claiming is free: crawl every page, re-audit as you fix, and track your score over time.

Sign in to claim

The fix queue

40 points sit between impeccable.style and 100: 9 open questions and 7 missing protocol checks, ordered by estimated payoff.

Point estimates are per fix under scoring v2. They are not additive to a promised total.

01limits · importance highGoes silent+7 content pts est.

We're migrating a massive legacy UI with thousands of variables. What's the upper limit on how many design tokens the Distill command can process in a single run before it errors out or truncates?

What the pages say

No page on the site addresses this.

The fix

Add a 'Limits' or 'Performance' section to the /impeccable distill documentation specifying the maximum number of design tokens or variables the command can process in a single run, or clarify if Distill does not process design tokens directly.

Where we looked: /docs/distill, /docs/distill, /docs/distill, /docs/distill

confidence high · grounding world-knowledge · weight 0.00 · Absent

02limits · importance highGoes silent+7 content pts est.

Overdrive seems like it could blow through my API budget fast. Is there a built-in cap on how many design variations it will generate per request, or can I set that threshold myself?

What the pages say

No page on the site addresses this.

The fix

Add a Limits, Pricing, or FAQ section to the Overdrive docs that explicitly states whether there is a built-in cap on design variations per request and whether users can configure that threshold.

Where we looked: /docs/overdrive, /docs/overdrive, /docs/overdrive, /docs/overdrive, /docs/overdrive, /docs/overdrive

confidence high · grounding synthesized · weight 0.00 · Absent

03features · importance mediumGoes silent+7 content pts est.

When I run Colorize on a light-mode UI, does it automatically generate a matching dark-mode palette, or do I have to run a separate command like Adapt to get that?

What the pages say

No page on the site addresses this.

Source: /docs/colorize, /docs/colorize, /docs/colorize, /docs/colorize, /docs/adapt

confidence high · grounding synthesized · weight 0.00 · Absent

04logging · importance lowGoes silent+7 content pts est.

When Clarify resolves an ambiguous prompt, does it keep a log of the alternative interpretations it rejected so I can review why it chose a specific layout direction?

What the pages say

The documentation does not mention whether Clarify keeps a log of alternative interpretations it rejected, nor does it describe any way to review why a specific direction was chosen.

The fix

Add a section to the Clarify documentation explaining whether the command logs alternative interpretations or rejected options, and how users can review its decision rationale.

Where we looked: /docs/clarify, /docs/clarify, /docs/clarify

confidence high · grounding synthesized · weight 0.00 · Absent

Hedged · 5 of 15 questions

A buyer gets something, then has to guess the rest. Expand any row for the evidence and the fix.

05I'm a solo designer using Figma — do I need to install anything locally to start using Impeccable, or is it entirely browser-based?getting-startedHedges+2 pts est.

I'm a solo designer using Figma — do I need to install anything locally to start using Impeccable, or is it entirely browser-based?

What the pages say

The chunks state that Impeccable includes an 'Optional CLI + Chrome extension' alongside the core agent skill, and the getting-started tutorial says to 'Install Impeccable.' However, none of the chunks mention Figma, and none explicitly state whether the core product is entirely browser-based or what a Figma user needs to install. The supported tools listed are AI coding assistants (Cursor, Claude Code, VS Code Copilot, etc.), not Figma.

The fix

Add a clear 'Getting Started' section that states what (if anything) needs to be installed locally, whether a browser-based workflow exists, and explicitly address Figma compatibility or redirect designers to the supported AI coding tools.

Where we looked: impeccable.style, /tutorials, /.well-known/webmcp.json

confidence medium · grounding synthesized · weight 0.65 · Vague

06We have an existing design system documented in Notion and PDF. Can Impeccable's Teach command ingest those directly, or do I need to convert everything to a specific structured format first?integrationHedges+2 pts est.

We have an existing design system documented in Notion and PDF. Can Impeccable's Teach command ingest those directly, or do I need to convert everything to a specific structured format first?

What the pages say

The documentation does not state that `/impeccable teach` can ingest Notion or PDF files directly. Instead, the retrieved chunks describe that Teach scans your codebase (README, package.json, components, tokens, brand assets) and runs a discovery interview to write `PRODUCT.md`. The companion `/impeccable document` command scans code tokens (CSS custom properties, Tailwind config, CSS-in-JS themes) to generate a `DESIGN.md` following the Google Stitch format. The site does not clarify whether existing external documentation must be manually converted to these structured files.

The fix

Add a FAQ or docs section explicitly stating whether Teach or Document can ingest external files (Notion, PDF, Figma exports) or if users must manually translate existing design systems into PRODUCT.md/DESIGN.md.

Where we looked: /docs/teach, /tutorials/getting-started, /docs/document

confidence medium · grounding synthesized · weight 0.65 · Vague

07The Animate command sounds useful for our micro-interactions, but does it actually let me specify spring physics parameters, or does it just spit out generic CSS ease-in-out values?technicalHedges+2 pts est.

The Animate command sounds useful for our micro-interactions, but does it actually let me specify spring physics parameters, or does it just spit out generic CSS ease-in-out values?

What the pages say

The retrieved chunks do not mention spring physics parameters for the Animate command. However, the docs show the command outputs specific named easing curves such as 'ease-out-quart' (e.g., 'Success screen enters with opacity + translateY(8px), 260ms, ease-out-quart') and animates only transform and opacity, rather than generic CSS ease-in-out values.

The fix

Clarify on the Animate docs page whether users can supply physics-based spring parameters (stiffness, damping, mass) or if the command only produces predefined CSS easing curves.

Where we looked: /docs/animate

confidence medium · grounding stated · weight 0.75 · Vague

08For the Live command, can multiple designers edit the same AI-generated prototype simultaneously with live cursors, or is it strictly single-user with read-only sharing?operationsHedges+2 pts est.

For the Live command, can multiple designers edit the same AI-generated prototype simultaneously with live cursors, or is it strictly single-user with read-only sharing?

What the pages say

The site does not mention multi-user editing, live cursors, or any sharing model for Live Mode. It describes Live Mode as a single-user workflow: one developer runs `/impeccable live` in their AI harness, picks an element in their local dev server, generates three variants, and accepts one to be written back to their source file. There is no discussion of simultaneous collaboration or read-only sharing.

The fix

Add explicit documentation on the Live Mode page clarifying whether multiple users can collaborate simultaneously with live cursors, or if sessions are strictly single-user.

Where we looked: /tutorials, /live-mode, /.well-known/webmcp.json, /docs/impeccable

confidence high · grounding synthesized · weight 0.70 · Vague

09The Optimize command mentions image compression — does it default to a specific quality percentage like 80%, or does it dynamically target a maximum file size in kilobytes?config-defaultsHedges+1 pt est.

The Optimize command mentions image compression — does it default to a specific quality percentage like 80%, or does it dynamically target a maximum file size in kilobytes?

What the pages say

The `/impeccable optimize` documentation says the command addresses images and assets by checking lazy loading, responsive images (`srcset`, `sizes`), modern formats (WebP, AVIF), and dimensions to prevent CLS, but it does not specify whether image compression defaults to a specific quality percentage or dynamically targets a maximum file size in kilobytes.

The fix

Add a configuration or behavior section to the `/impeccable optimize` docs that explicitly states whether image compression uses a default quality threshold (e.g., 80%), targets a max file size in KB, or relies only on format conversion without re-quantization.

Where we looked: /docs/optimize

confidence high · grounding stated · weight 0.80 · Vague

What the pages already answer cleanly

6 of 15. Answered with enough specificity for a buyer to act on. This is the bar the queue above is chasing.

complianceAnswered

We need to meet WCAG 2.1 AA standards. Does the Audit command specifically check contrast ratios and focus states for accessibility compliance, or is it only looking for visual brand inconsistencies?

What the pages say

The `/impeccable audit` command specifically checks both contrast ratios and focus states as part of its Accessibility dimension. The docs state that the Accessibility scan includes "WCAG contrast, ARIA, keyboard nav, semantic HTML, form labels," and sample output shows findings such as "Contrast 3.1:1 on disabled button state" and "No visible focus indicator on custom dropdown." It is not limited to visual brand inconsistencies; Accessibility is one of five distinct dimensions scored separately from Theming.

Source: /docs/audit, /docs/audit

confidence high · grounding synthesized · weight 0.95 · Answered

securityAnswered

My agency has strict client NDAs. Does Impeccable's privacy policy explicitly state that our uploaded design files won't be used to train third-party AI models?

What the pages say

No. Impeccable's privacy policy does not explicitly state that uploaded design files won't be used to train third-party AI models. The policy says Impeccable is 'an open-source collection of agent skills (text files) that run locally in your AI coding tool' and that 'the skills themselves collect no data, make no network requests, and have no analytics.'

Source: /privacy

confidence high · grounding stated · weight 1.00 · Answered

product-overviewAnswered

How many design commands are currently available in Impeccable's core library? I saw a long list but want to know the exact number so I can plan our team onboarding.

What the pages say

Impeccable currently offers 23 design commands in its core library.

Source: impeccable.style, /docs, /llms.txt

confidence high · grounding stated · weight 1.00 · Answered

integrationAnswered

If I use the Extract command to pull assets from a design, does it export to SVG and PNG only, or does it also support PDF and Figma's .fig format?

What the pages say

No. The `/impeccable extract` command does not export assets to SVG, PNG, PDF, or Figma's .fig format. It is a code-oriented tool that pulls reusable components, tokens, and patterns from your codebase into a design system—extracting repeated CSS values, component variants, spacing scales, and text styles into tokens and component APIs, then migrating call sites.

Source: /docs/extract, /docs/extract, /docs/extract

confidence high · grounding synthesized · weight 0.95 · Answered

semanticsAnswered

I'm confused by the name — is Harden meant for stress-testing design system components against edge-case content, or is it about securing the actual frontend code against vulnerabilities?

What the pages say

Harden is meant for stress-testing design system components against edge-case content and real-world conditions — not for securing frontend code against vulnerabilities. The site describes `/impeccable harden` as the step for "the day your interface meets reality," handling messy real user data such as 60-character names, German translations, prices in the billions, 500 errors, offline modes, and right-to-left text. It works through four dimensions of real-world resilience: (1) text and data extremes, (2) error scenarios, (3) internationalization, and (4) device and context, applying concrete fixes like overflow handling, informative error UI, and i18n-safe layouts.

Source: /docs/harden, /docs/harden

confidence high · grounding synthesized · weight 0.95 · Answered

methodologyAnswered

The Slop page talks about low-quality AI design output, but does it actually provide a concrete checklist or numerical scoring rubric we can use to flag slop in our own workflows?

What the pages say

The Slop page provides a concrete checklist of 37 specific AI slop patterns organized into categories such as Visual Details, Typography, Color & Contrast, Layout & Space, Motion, Interaction, Responsive, and General Quality. Each rule indicates its detection method—CLI, Browser, or LLM only—and the page shows concrete examples like "Modal Abuse" (complex settings crammed into a modal). However, the Slop page does not provide a numerical scoring rubric; numerical scores (e.g., heuristic scores from 0–4 and cognitive-load failure counts out of 8) are part of the separate `/impeccable critique` tool.

Source: /slop, /slop, /docs/critique

confidence high · grounding synthesized · weight 0.95 · Answered

Protocol plumbing · 56/1009 of 16 checks pass · each fix +6 protocol pts est.

The other half of the score: 16 checks for the files and headers agents look for. The 7 below are installs, not judgment calls, and most are an afternoon. Expand any for the snippet and the standard it follows. They sit after the queue because none of them changes what your pages say.

llms.txtDiscoverability+6 pts est.

Sitedex generates this file from your crawl. Grab it in Files from this audit below.

Standardllmstxt.orgCommunity spec

AI crawler accessAccess+6 pts est.
Install snippet
User-agent: GPTBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: anthropic-ai
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: Googlebot-Extended
Allow: /

StandardRFC 9309IETF RFC

Markdown negotiationRendering+6 pts est.

StandardRFC 9110 + 7763IETF RFC

MCP cardInteraction+6 pts est.

Sitedex generates this file from your crawl. Grab it in Files from this audit below.

StandardModel Context ProtocolCommunity spec

OpenAPI specInteraction+6 pts est.

StandardOpenAPI SpecIndustry standard

WebMCP widgetInteraction+6 pts est.

Sitedex generates this file from your crawl. Grab it in Files from this audit below.

StandardW3C WebMCP draftW3C / WHATWG

Organization schemaIdentity+6 pts est.

Sitedex generates this file from your crawl. Grab it in Files from this audit below.

StandardSchema.org + JSON-LDIndustry standard

Already passing 9 of 16: robots.txt, sitemap.xml, Content signal, Clean crawl, Server-rendered content, Canonical URLs, Meta descriptions, HTML lang attribute, Sitemap lastmod.

Ask this site’s index

Sitedex already serves impeccable.style as an MCP endpoint. Ask impeccable.style anything an AI agent might ask, and see what its index returns. (To score your own site, use the form below.)

Snippets & configs

For developers and the engineer-on-call: copy these into your tools or your site.

Files from this audit

Built from this crawl. Download or copy each, then install it at the path noted.

llms.txt

Built from this crawl. Install at /llms.txt so agents start here.

organization.json

Organization JSON-LD, pre-filled from this crawl. Wrap in a ld+json script.

server-card.json

MCP server card built from this crawl. Host at /.well-known/mcp/server-card.json.

webmcp.json

WebMCP discovery manifest built from this crawl. Host at /.well-known/webmcp.json.

MCP endpoint

https://mcp.sitedex.dev/s/impeccable-style/mcp

The URL anyone's agent points at. Read-only; safe to share.

Claude Code

claude mcp add impeccable --transport http https://mcp.sitedex.dev/s/impeccable-style/mcp

One command, then the agent has it.

Cursor / Continue

{
  "mcpServers": {
    "impeccable": {
      "url": "https://mcp.sitedex.dev/s/impeccable-style/mcp"
    }
  }
}

Drop into mcp.json.

WebMCP: two parts

WebMCP-capable browsers run the widget at runtime. Crawlers without JS rendering need the discovery manifest to find your tool surface. Install both.

1 · Widget script

<script async src="https://sitedex.dev/widget.js"></script>

Drop in <head>. WebMCP-capable browsers (Chrome 146+ Origin Trial) call navigator.modelContext.provideContext() via this script.

2 · Discovery manifest

{
  "$schema": "https://wellknownmcp.org/schemas/webmcp.json",
  "name": "impeccable.style",
  "tools": [
    { "name": "search", "description": "Search impeccable.style's indexed content." },
    { "name": "get_page", "description": "Fetch a page from impeccable.style as markdown." }
  ]
}

Host alongside the script at /.well-known/webmcp.json. Crawlers that don't render JS rely on this.

Your turn

See which of these questions your site goes silent on.

Free, about 5 minutes. We crawl your site, test it against the buyer questions your category asks, and name what’s vague, contradictory, or missing, plus the files AI agents look for.

ComingEmbeddable grade badgeScore history and deltasOpt-in public board