Impeccable
pbakaus/impeccable
A design skill with 24 commands and a detector for the common tells of AI-generated frontend design.
Impeccable is a free, Apache-2.0 design skill for AI coding agents, created by Paul Bakaus. It is one skill with 24 commands, such as audit, critique and polish, plus 61 detector rules that find common faults in AI-generated interfaces without calling a model. It supports 17 tools including Claude Code, Cursor, Codex CLI, Gemini CLI and GitHub Copilot, and installs with npx impeccable install.
Facts checked on against the sources listed at the end of this page.
- GitHub stars
- 74,141
- Skills in the pack
- 1
- Agents supported
- 17named in the README
- Instruction size
- 1.6k wordsacross its SKILL.md file
How to install Impeccable
| Agent | Command | Copy |
|---|---|---|
| All supported agents | Run it from your project root. It detects your agents and asks whether to install for the project or globally. Then run /impeccable init in your agent. | |
| Claude Code plugin | Then open /plugin and install Impeccable from the list. | |
| GitHub Copilot in VS Code | Needs VS Code 1.109.3 or newer. This extension installs the skill without the automatic hooks. | |
| Codex | After npx impeccable install, open /hooks in Codex and approve the project hook. | |
| Grok Build | |
Commands for 1 more agent
| Updating | |
|---|
Commands as given in the README on 2 October 2026.
What is inside Impeccable
| Skill or command | What it makes the agent do |
|---|---|
| /impeccable init | One-time setup. Inspects the project, asks about gaps and writes PRODUCT.md with the audience, purpose and constraints. |
| /impeccable craft | The full flow: shape the design, then build it with visual iteration. |
| /impeccable critique | A UX design review covering hierarchy, clarity and emotional resonance. |
| /impeccable audit | Technical quality checks for accessibility, performance and responsive behavior. |
| /impeccable polish | A final pass for design system alignment and shipping readiness. |
| /impeccable typeset | Fixes font choices, hierarchy and sizing. |
| /impeccable layout | Fixes layout, spacing and visual rhythm. |
| /impeccable harden | Adds error handling, internationalization, text overflow handling and edge cases. |
| /impeccable live | Visual variant mode: iterate on elements in the browser against a local development server. |
One skill with 24 commands, all run through /impeccable. Use /impeccable pin to turn a command into its own shortcut, for example /audit.
What Impeccable runs on your machine
| Hooks | Yes. On Claude Code, GitHub Copilot, Codex, Cursor and Grok Build the installer adds a hook that runs the design detector when the agent edits UI files and reports findings back to the agent. |
|---|---|
| Bundled scripts | Yes. Each skill copy ships a launcher that runs the Impeccable engine, a self-contained binary that sits next to the launcher or is downloaded once into ~/.impeccable/bin/. |
| Network calls | Yes. The launcher downloads the engine binary on first run if it is not already present. Live mode runs a helper on localhost against your local checkout. |
| API keys needed | None. The README says the detector rules run with no LLM and no API key. |
These are facts read from the repo, not a security rating. Read the files before you install any skill.
How much context Impeccable uses
| The skill | 1,602 words in its SKILL.md file |
|---|---|
| Loaded at start | Only each skill's name and description. The full text loads when a task needs that skill. |
Word counts are measured from the repo every day. Reference files a skill loads on demand are not counted.
How Impeccable works
Impeccable started from Anthropic's frontend-design skill. Its premise is that models trained on the same templates repeat the same choices: Inter for everything, purple-to-blue gradients, cards nested in cards and gray text on colored backgrounds. The skill gives the agent explicit guidance on what to avoid and a shared vocabulary of commands such as bolder, quieter and distill.
The detector is separate from the model. A command line tool and a browser extension run the 61 deterministic rules on a folder, an HTML file or a live URL. As you run commands, Impeccable writes working files under a .impeccable folder in the project, and the README lists which of them to keep out of git.
A good fit for
- Frontend work where agent output looks like every other generated site
- Reviewing an existing interface with audit and critique before changing it
- Teams that want design checks to run automatically when the agent edits UI files
- Checking a page in CI or by hand with the detector, with no model involved
Skip it if
- Backend or non-visual work: every command is about interfaces
- Setups that do not allow hooks or downloaded binaries, since the hook can fetch and cache the engine on first run
- Editing a deployed production site in live mode, which the README says is not supported
- Cursor users on the stable channel, if installing by manual copy: the README says Cursor skills need the Nightly channel with Agent Skills enabled
Is Impeccable still maintained?
| Last commit | , 16 hours ago |
|---|---|
| Latest release | skill-v4.5.0, 2 October 2026 |
| Stars gained in 7 days | Shown after 7 days of tracking |
| Stars gained in 30 days | Shown after 30 days of tracking |
| Open issues and pull requests | 54 |
| Forks | 4,469 |
Tools Impeccable works with
Questions about Impeccable
How do I install Impeccable?
From your project root run npx impeccable install, then run /impeccable init inside your AI coding tool. The installer detects which agents you have.
Which tools does Impeccable support?
The README lists 17: Cursor, Claude Code, GitHub Copilot, DeepSeek Harness, Gemini CLI, Codex CLI, Grok Build, Hermes Agent, OpenCode, Pi, Kiro, Trae, Rovo Dev, Qoder, Mistral Vibe, Veto and Google Antigravity.
Is Impeccable safe to install?
It is open source under Apache-2.0. It installs a hook on five agents and runs a binary that may be downloaded on first use, so read the hook section of the README before unattended runs. In Claude Code the hook runs independently of model tool approval.
Does Impeccable need an API key?
No. The skill uses your agent's own model, and the 61 detector rules run with no LLM and no API key.
Is Impeccable free?
Yes. It is open source under the Apache-2.0 license.