@pbakaus: Impeccable 4.1 is here: core: -critique results more reliably prints in chat before asking you questions - the design h…
Summary
Impeccable 4.1 releases with 43 PRs, improving critique reliability, design-system consistency, native app review, plugin installs, and live mode support across more dev environments.
View Cached Full Text
Cached at: 08/14/26, 03:36 PM
Impeccable 4.1 is here:
core: -critique results more reliably prints in chat before asking you questions
- the design hook nags less, and ignores can be recorded once
- reduced-motion guidance restored
greenfield design:
- switch between comps of your page/app before any code or skip mockups and go straight to code
- steer ideation to be bolder or safer in the interactive question tool
- mockups for new pages of an existing site/app stay more consistent
- much better fallback for agents without image generation
native apps:
- ios and android builds get reviewed against native conventions
- polish collects native evidence instead of asking for browser screenshots
- web-only checks stop firing on native code
platforms and setup:
- windows: installs, the decision page, and the roll all work now
- plugin installs: the four subagents actually load (they silently never did)
- support for @NousResearch hermes agent and @antigravity
design issue detection:
- no false low-contrast on gradient backgrounds, image backgrounds, or covered text
- data-impeccable-ignore silences only the element you put it on
- broken-image warnings stop firing on commented-out code
- blade files show up in directory scans
- paths with spaces or quotes stop breaking commands
live mode:
- works on ddev, valet, and other non-localhost dev hosts
- svelte variants render again (broke due to svelte upstream changes)
- copy edits survive long pages
43 prs since the last release - great time to upgrade or try out the latest: npx impeccable install
Impeccable: The missing design vocabulary for agents
Source: https://impeccable.style/
Detects and removes AI slop.
59 checks catch the defaults agents keep reaching for before they become your design.
23 commands give you a shared design vocabulary with your AI. Each names one kind of intervention:/typesetfor type,/distillfor complexity,/auditfor production quality. Pick any to see it in action, orbrowse the full reference →
Give direction in chat. Refine the result on the page.
Every change lands in your source.
mainClaudex Cocursor
Agent threadWorking treeYou
/impeccable polishthe pricing page. Keep our sharp corners and sober palette. Remove the AI tells.
Agent
I’ll polish the page within its existing design system, then run the detector before I hand it back.
DESIGN.md loadedExisting system preserved
4 tells foundAI beige · italic serif · side-tab · pulsing dot
Written to source0 detector findings
Agent
Starting Live Mode against the running app. I’ll open the page, then wait for element picks and page-level steering.
Live helper startedPicker injected into the app
localhost:3000 openedProduction source connected
Waiting for inputPick an element or steer the whole page
/impeccableChoose a command
BrowserLive previewlocalhost:3000
Generating variants…
1 / 3
Variant 3 written to source
Ask a model to be creative and it builds its favorite idea, every run. So Impeccable deals challengers: human-reviewed worlds, each a complete graphic system with laws of its own. They compete against the ideas the model derives from your product, and the direction that wins takes the build, fully committed. The hand below is dealt live from the 177 highest-rated worlds in the deck by the same roll API the skill uses.Why dice? Read the research →
Psst…the deck grows every week in Pro →
01### Respects your design system.
Inherits your tokens, components, and conventions instead of overwriting them. New work fits the system, not the model’s defaults.
›Walking src/…
✓src/styles/tokens.css14 colors, 6 font sizes, 7 spacing steps
✓tailwind.config.tstheme.extend merged
✓src/components/34 components
· Button: primary, secondary, ghost, destructive
· Card: 3 variants, 8px / 16px radii
· Input, Select, Dialog, Toast, Table…
✓DESIGN.mdbrand rules loaded
→Matching your system. Not inventing one.
02### Real product context.
Captures the brief once, inPRODUCT\.md. Every command reads it before designing.
UsersSREs on call, reading fast, often in the dark.
ModeOperate. Design serves the task.
Brand voiceCalm, clinical, no hype.
Anti-referencesPurple gradients. Glassmorphism. “Boost your productivity.”
03### Travels as DESIGN.md.
/impeccable documentwrites one in theGoogle Stitch format. Your visual system becomes portable.
WordmarkSpaced 400
IM
Avenir Next
BodyRegular 400
Aa
Avenir Next
Components34
PrimaryGhost
Email address
Readable by every DESIGN.md-aware tool.
04### Knows what the surface is for.
A landing page should persuade. A dashboard should help someone work. Impeccable names the mode before it designs.
ExperienceLet the work lead
05### Block slop before it ships.
Hooks inspect each UI edit and send findings straight back to the agent. The same detector can fail a PR.
1Edit landsAgent updates the hero.
2Hook returns two findingsAI beige · clipped label
✓Agent corrects the fileSecond pass: clean
Deep stop audit0 findings · PR ready
Get started
Install via
On the GitHub Copilot app? Impeccable isbuilt in. Skip the install and enable it underSettings → Experimental.
npx impeccable install
Installs the build tailored to your harness and model. RequiresNode 22.12+.
Why one command, many buildsimpeccableClaude Codehooks + subagentCursorpre-edit hookGemini CLI+ Gemini-tuned slop rulesCodex CLI+ Codex-tuned slop rulesSame skill, recompiled per target. The part that matters: builds for models with known tells carry extra slop rules banning that model’s habits. The Gemini build kills its image-on-hover motion; the Codex build refuses ghost-cards and over-rounding. Native hook wiring and a subagent ship too, where the tool supports them.
First run · in your agent chat
/impeccable init
Works with Cursor, Claude Code, GitHub Copilot, Gemini CLI, Codex CLI, Grok Build, and every other major AI coding harness.
Read the setup guide·See what’s new·Got questions?
Learn more:Getting started·Designing with Impeccable·command docs
Similar Articles
Impeccable: Design skills for AI harnesses
Impeccable is a suite of 18 CLI commands, a Chrome extension and library that embeds design-quality checks into AI coding workflows to detect and fix common UI anti-patterns without needing an LLM.
@danshipper: vibe check: Opus 4.7 feels like it's gotten a lot better recently. Both at coding and writing / strategy / deep thinkin…
Users report noticeable improvements in Opus 4.7's performance for coding, writing, and strategic reasoning tasks.
@bentossell: wait… if most people think 5.5 is better than 4.7, i assume that’s due to terminal coding benchmark… 4.8 is still outpe…
The tweet discusses the release of Claude Opus 4.8, which improves upon Opus 4.7 with sharper judgment and longer independent work, though it notes that version 5.5 still outperforms it on a terminal coding benchmark.
@github: .@impeccable_ai earned its way to 40k+ stars on GitHub by delivering exceptional value. Today, we are making design and…
Impeccable AI is now a built-in skill in GitHub Copilot, making design and quality a built-in layer for all creators.
@RayFernando1337: The bugs that cause churn almost never show up in a diff, and you only really catch them when you stop reviewing code a…
A developer shares a workflow using Cursor's Opus 4.8 Max Thinking model with subagent harness, and introduces a GitHub repository with installable skill files for AI coding agents, including a 'running-bug-review-board' skill that performs live QA testing.