What It Is

A reusable evaluation skill packaged for Claude Code that takes any website URL and runs it through Jakob Nielsen’s ten usability heuristics — visibility of system status, match between system and the real world, user control and freedom, consistency and standards, error prevention, recognition over recall, flexibility and efficiency, aesthetic and minimalist design, error recovery, and help / documentation.

The output is a professional PDF audit report with executive summary, per-heuristic findings (Pass / Partial / Fail), evidence-based observations, and a priority matrix ranking issues from Critical to Cosmetic.

→ github.com/mertony8/claude-skills/tree/main/ux-heuristic-evaluator

How It Works

Five phases the skill walks through, end to end:

Why I Built It

Heuristic evaluations are the most reliable cheap-and-fast quality check in the design toolkit, and the easiest one to skip when the schedule tightens. Encoding the method as a skill removes that excuse: any reviewer (designer, PM, engineer) can run a first pass without booking a workshop or hiring a consultant.

It’s also the simplest possible argument for what a “skill” is when software does the work. The method is well-defined; the steps are mechanical; the variability is in the evidence and the writing. That’s the shape of work agents handle well — but only if the policy around them (scope, crawl strategy, evidence collection, output format) is encoded explicitly.

Sample Reports

The repo includes real reports generated by the skill against public sites:

The PDFs live in examples/ in the repo.

Requirements

Status

v1 shipped, open-source. Iterating as new sites get evaluated and the report format gets pressure-tested. Issues, PRs, and “run this on my site” requests welcome on GitHub.