Web Interface Guidelines Audit

This skill reviews UI source against a living checklist of interface best practices and reports what falls short. Because the rules are fetched fresh each run, the audit reflects the most recent published guidance rather than a snapshot frozen into this file.

What it does

Given one or more files (or a glob), it retrieves the current guidelines, reads the target code, evaluates every applicable rule, and emits concise findings keyed to file:line so each issue is easy to locate and fix.

Procedure

1. Pull the current rules. Fetch the guidelines document from the source below at the

start of every review — do not work from memory or a cached copy. The fetched content carries both the rules to apply and the exact output format to follow.

1. Gather the targets. Read the files named in the request. If none were given, ask

which files or pattern to review before continuing.

1. Evaluate. Walk the code against each rule in the fetched guidelines.
2. Report. Write findings in the terse file:line form the guidelines specify, one

issue per line.

Where the rules come from

Retrieve the latest rules before each audit:

https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md

Use a web fetch to load that URL. Treat whatever it returns as the authority for both the checks and the reporting format — it supersedes anything described here.

How this fits a design workflow

Two skills bracket the act of building UI: one informs the design before code exists, this one inspects the result after.

1. DESIGN  -> absorb the principles
2. BUILD   -> implement the interface
3. AUDIT   -> run this review        <- you are here
4. FIX     -> resolve the findings