What it does

Impeccable is a “skill” — essentially a curated prompt bundle — that installs into AI coding tools (Cursor, Claude Code, Copilot, Gemini CLI, and nine others). It gives you 23 slash commands like /impeccable polish, /impeccable bolder, and /impeccable distill that invoke domain-specific design guidance. A standalone CLI can also scan code or URLs for 24 deterministic anti-patterns without calling an LLM at all.

The interesting bit

The project treats AI design taste as a supply-chain problem. Models trained on the same SaaS templates keep outputting the same tells: Inter everywhere, cards nested in cards, gray text on colored backgrounds. Impeccable loads seven reference files (typography, color, motion, spatial, interaction, responsive, UX writing) on every command, plus a brand-vs-product register that adjusts defaults. It is essentially a structured argument against the lowest-common-denominator output of frontier models.

Key highlights

• 23 commands covering the full lifecycle: init, shape, craft, critique, audit, polish, harden, live
• 27 anti-pattern rules with explicit negatives: no pure black/gray, no bounce easing, no nested cards
• CLI detector runs regex-only with --fast --json, no API key required
• Auto-detects your AI harness and installs the right build via npx impeccable skills install
• Pinnable shortcuts: /impeccable pin audit creates a standalone /audit command

Caveats

• Several supported tools require manual beta setup (Cursor Nightly, Gemini CLI preview, Trae’s two separate config directories)
• The “browser extension” mentioned in the anti-patterns section is referenced but not detailed in the README
• 33,833 stars with no GitHub topics listed, which is slightly odd for a project of that size

Verdict

Worth installing if you are tired of manually correcting the same AI-generated design clichés. Skip it if you already maintain your own design-system prompts or do not use any of the supported AI coding tools.