Review website code with AI
Review a focused frontend change in ordered passes and receive severity-ranked findings with a location, a concrete problem and a suggested fix.
Review website code with AI through a sequence that puts behaviour ahead of polish. The taste-code-review skill asks for findings with severity, file location, a concise problem and a practical fix, so the review can lead to action.
Its passes cover behaviour, structure, consistency, defensive security, performance and accessibility. Empty states, repeated actions and error paths belong in the review alongside the successful flow. Keep the change focused and provide the surrounding conventions and tests; an isolated screenshot cannot establish what the code does. The resulting report helps you decide what to fix, but it does not independently authorise a merge or replace testing in the project.
What it does
- Check successful and failing interaction paths.
- Review file structure and responsibilities.
- Separate blockers from smaller suggestions.
- Give each finding a location and a fix.
- Read the relevant tests with the change.
Install and use
- Install the skill with its core guide in a compatible coding host. Provide the diff, relevant conventions and tests, keeping credentials out of the material.
- Ask: “Review the signup-form change, including empty values, repeated clicks and error messages. Report actionable findings with severity and locations.” Define the scope.
- Assess the findings before applying fixes, then rerun project checks and inspect the affected behaviour. Keep subjective nits distinct from issues that should block the change.
FAQ
Source
Made by Mustafa. Original skill: taste-code-review, published on GitHub under the MIT licence.
View on GitHub