audit-website: Claude Code Skill & Agent Tool
An open-source skill by squirrelscan (opens in a new tab), published in the squirrelscan/skills (opens in a new tab) repository on GitHub. Runs the squirrelscan CLI across 300+ rules — SEO, performance, security, accessibility and content — then maps each issue to your source files, applies fixes and re-audits until the site scores well.
Install audit-website
Requires Node.jsnpx skills add squirrelscan/skills --skill audit-websiteRun this command in your project terminal to install this skill for Claude Code.
What this skill does
audit-website connects your AI agent to squirrelscan, a website auditing CLI, and then makes the agent fix what it finds. A single scan runs more than 300 rules across SEO, performance, security, technical health, content, accessibility and over a dozen other categories, and returns a report written to be read by a language model, not just by a person.
What sets it apart is the fix loop. The agent presents the score, grade and top issues by severity, proposes the fixes it can make and asks for your approval, then maps each finding to the template, component or content file that causes it. It applies the approved fixes in batches, checks that the project still builds, re-audits and shows before-and-after scores, repeating until the target is met. Errors come first, then high-rank warnings; judgment calls are flagged for your review instead of guessed.
The target score depends on the starting score, and the final sign-off uses a full crawl, because the quick pass only samples part of the site. It needs the squirrel CLI on your PATH. Use it with Claude Code to bring a site to a high audit score.
Key Capabilities
- 300+ rules across SEO, performance, security, accessibility, content and more
- Maps each issue to the template, component or content file behind it
- Proposes fixes for your approval, applies them in batches and checks the build
- Re-audits with before-and-after scores until the target score is reached
- Signs off on a full crawl; judgment calls are flagged, never guessed
How to use with Claude Code & AI Agents
Install the skill
Run the command above at the root of your project. The CLI asks which agents to set up — Claude Code, Cursor and other supported agents — and adds audit-website to the project, so teammates who pull the repository get it too. Add-gto install it for your user account across every project.Open a new agent session
Start a fresh Claude Code session in that project so the agent loads the new skill. Have ready: squirrel CLI on your PATH.Ask in plain words
Describe the task, for example:“Audit my site with squirrelscan and fix the issues until we score above 90.”The agent matches your request to the skill's description and follows its SKILL.md instructions step by step. Read them on GitHub before using the skill on a system that matters.