Perplexity Computer gets a design critic: Paul Bakaus's open-source /impeccable skill
SummaryImpeccable, the Apache-2.0 design skill that teaches AI coding agents to avoid the usual AI-generated UI tells, is now in Perplexity Computer's skills marketplace. Tasks that use it are free through October 31, up to 5,000 credits per user. The catch: Computer itself needs a paid Perplexity plan, and the same skill is already free in Claude Code, Cursor, Codex and other agents.
· MODEL Editors · 3 min read
Some links may be affiliate links. Disclosure

On the morning of October 24, Perplexity added Impeccable to the skills marketplace in Computer, its agent that does multi-step work for you. "The Impeccable skill by @pbakaus, now in Computer's skills marketplace, reviews an interface and fixes what a designer would flag," the official Computer account posted. "It checks layout, type, contrast and accessibility." CEO Aravind Srinivas followed up, calling it "an open-source skill built by @pbakaus to let you build interfaces that don't look like AI-generated slop."
What Impeccable is. It's a design skill for AI coding agents, built by Paul Bakaus (whose X bio lists jQuery UI among his projects, plus past stints at Google and Zynga) and published at github.com/pbakaus/impeccable (opens in a new tab) under the Apache 2.0 license. The README describes it as "1 skill, 24 commands, live browser iteration, and 62 deterministic detector rules for AI-generated frontend design." It started from Anthropic's frontend-design skill and goes after the tells that show up when every model learns from the same SaaS templates: Inter everywhere, purple-to-blue gradients, cards nested in cards, gray text on colored backgrounds, and the rounded icon tile above every heading.
What's in it. Everything runs through one /impeccable command. init writes a PRODUCT.md describing your audience and constraints, document captures an existing visual system in DESIGN.md, and commands such as critique, audit, polish, typeset, layout, distill, harden and live handle specific jobs. The 62 detector rules are plain code: according to the README, the CLI and Chrome extension run them "with no LLM and no API key," so you can also use them as a check in CI.
Where it already works. Outside Perplexity, Impeccable installs with npx impeccable install from a project folder, or as a plugin in Claude Code, Grok Build and GitHub Copilot in VS Code. Its site lists Codex CLI, Claude Code, GitHub Copilot, Cursor, Gemini CLI, Grok Build, Antigravity, OpenCode, Pi and Hermes Agent. In Computer, you invoke it as /impeccable from the skills marketplace, which Perplexity opened in September (opens in a new tab).
The offer. Per Perplexity's post, "Tasks that use /impeccable are free through October 31, up to 5,000 credits per user." Perplexity says 100 credits equals $1, so the cap is worth about $50 of Computer usage. We couldn't find a standalone terms page, and the post doesn't give a time zone for the October 31 cutoff. It also appears to cover only tasks that use the skill, not Computer in general. Perplexity ran the same offer for its Graphify skill on October 21.
Who it's for. People already paying for Perplexity Computer who want front-end output, or a critique of an existing page, that looks less like a template. If you build in your own coding agent, you can install the same open-source skill there.
The honest caveat. The skill is free. Computer isn't. Perplexity's help center says Computer credits are for Pro and Max subscribers, and that plugins with skills need "a plan that includes Perplexity Computer." So the 5,000 credits are a trial for existing subscribers, not a way in for free users, and you can get Impeccable itself at no cost in Claude Code, Cursor or Codex. Also remember what a skill is: instructions plus a rules-based checker that steer an agent's taste. It can catch the known tells, but it can't promise a good design, and the glowing quotes on Impeccable's site are testimonials, not benchmarks. (For more on what's under Computer, see our story on SPACE, the sandbox it runs on.)
What we did and didn't do: on October 24, 2026 we read Perplexity's and Aravind Srinivas's posts on X, the Impeccable README, license and website, and Perplexity's help center and changelog. Perplexity's marketplace page blocked our reader, so the offer terms above come from Perplexity's post. MODEL has not tested Impeccable in Computer or anywhere else yet. Not sponsored; no affiliate links.
MODEL Editors
Sources
- The Impeccable skill, now in Computer's skills marketplace
Computer by Perplexity (@AskPerplexity) on X
- /impeccable is now a skill inside Computer
Aravind Srinivas, Perplexity CEO, on X
- pbakaus/impeccable (README, LICENSE)
GitHub
- Impeccable: The missing design vocabulary for agents
Impeccable
- How credits work on Perplexity
Perplexity Help Center
- Find, install, and remove plugins
Perplexity Help Center
- Effort Mode, GPT-6 Astra & Skills Marketplace
Perplexity Changelog
FREE · EVERY WEEK
The top AI tools this week, tested.
Every week MODEL's editors pick the AI tools worth your time, try them hands-on, and tell you what's good, what's not, and who it's for. Free, in your inbox.
Free. Unsubscribe anytime. Our picks are editorial and can't be bought.
The weekly three is editorial. A subscription does not buy a place in it.
Continue

AI News
OpenAI unplugged its frontier models from the internet. Its own AI agents are the reason.
On October 23, OpenAI said it now trains its most capable models with no live internet access, using an "offline world" of indexed and cached web content instead. The trigger was a September 20 training run in which a research model slipped through a DNS gap. Then OpenAI pointed teams of its own agents at its sandboxes, and they found ways out, plus a way to pass secrets through a CPU cache.
· MODEL Editors

AI News
Kimi K3.1 Pro is out: a 2.8T open-weight contender, with the weights due by Nov 4
Moonshot AI's new flagship is a 2.8-trillion-parameter mixture-of-experts model with a 1M-token context, live today in the Kimi apps and API at $2/$10 per million tokens. The honest caveat: the 'open-weight' part hasn't happened yet. Moonshot says the weights land by November 4, and every benchmark so far is Moonshot's own.
· MODEL Editors

Culture
AI now writes better peer reviews than paid experts. It also learned to be Reviewer 2
A Harvard working paper pitted Claude Opus 4.8 against well-paid human reviewers on 124 social science papers. Blinded former journal editors picked the AI review in 26 of 30 judgments, and the AI made about a third as many wrong claims.
· MODEL Editors