I Built a Bot That Audits Websites and Emails Me the Report

Auditing websites is a big part of what I do. It's also, if I'm honest, a bit repetitive. Run the performance tools, check the Core Web Vitals, dig through the codebase, scan the dependencies for known vulnerabilities, write it all up in a report a client can actually read. Useful work, but a lot of it is the same steps over and over.

So I did the developer thing and automated it. Now a bot does the boring bit and emails me a finished report. Here's how it hangs together.

The idea

I wanted something that could:

1. Take a website (and its codebase).
2. Check the performance and Core Web Vitals.
3. Scan it for security issues and dodgy dependencies.
4. Turn all that into a proper, branded report - Word doc and PDF.
5. Email it to me, done, without me lifting a finger.

Basically, my own audit process, minus me sitting there doing it.

The stack

Two main pieces doing the heavy lifting:

• GitHub Actions - the engine. It runs on a schedule (or when I trigger it), spins up a runner, and does all the actual checking: a Lighthouse run for performance and Web Vitals, a dependency and vulnerability scan over the codebase, and a few extra checks I'd normally do by hand.
• Claude skills - the brains. Once the raw data's been gathered, a Claude skill reads through it, works out what actually matters, and writes it up in plain English. Not just "here's a number", but "here's what's wrong, why it matters, and what to do about it" - the bit that usually takes me the longest.

The nice thing about skills is they're repeatable. I've effectively taught it how I write an audit, so the output sounds like my reports rather than a wall of raw tool output.

How a run actually goes

1. The workflow kicks off and pulls down the site and its code.
2. It runs the performance pass (Lighthouse / Core Web Vitals) and the security and dependency scan, and dumps the results out as structured data.
3. Claude takes that lot, prioritises it, and writes the findings - quick wins, bigger jobs, and the why behind each one.
4. It drops the writeup into a templated Word document (my branding, my layout) and exports a PDF alongside it.
5. The bot emails both over to me, attachments and all.

By the time I've had a coffee, there's a finished audit sitting in my inbox that I can review, tweak if needed, and send on.

Why bother

A few reasons, beyond it being a fun thing to build:

• Consistency. Every audit covers the same ground in the same format. Nothing gets forgotten because I was in a rush.
• Speed. The grind that used to eat an afternoon now happens while I'm doing something else.
• I stay in the loop. This isn't "AI does my job". It does the legwork and the first draft. I still review every report before it goes anywhere near a client - the judgement's still mine, the robot just saves me the donkey work.

Where it's going

Right now it's pointed at my own projects and a watchlist of sites. The obvious next step is running it for clients on a schedule - a monthly health report that lands automatically, flagging anything that's drifted before it becomes a problem. Which, conveniently, is exactly what a good support and maintenance plan should be doing anyway.

This is the sort of thing I love about the job: spot the repetitive bit, build something to handle it, and free yourself up for the work that actually needs a human. If you've got a process in your own business that feels like it should be automated, have a chat with me - there's usually a way.