Prompt Check — prompt linter by Cairden Digital

Most bad AI answers are caused by the prompt, not the model. Prompt Check reads your draft before you hit send and tells you exactly what you left out.

It scores your prompt out of 100 and flags specific, fixable problems: missing output shape, no uncertainty rule, unbounded scope, judgement without reasoning, no context or constraints, several unrelated asks crammed into one message, vague quantities and dangling pronouns.

Every finding comes with a one-click fix that appends precise wording to your prompt.

When several things are missing at once, Restructure rewrites the whole draft into labelled sections (Context, Task, Audience, Constraints, Output), keeping your own wording untouched and flagging placeholders until you fill them. One click undoes it.

It also includes ten reusable instruction blocks that close common prompting failure modes: No fabrication, Source vs inference, Flag uncertainty, Reasoning then answer, Assumptions first, Tight output, Show the gaps, Argue against it, Check the numbers, and Ask before assuming.

Works on ChatGPT, Claude, Gemini, Perplexity and Le Chat. Alt+Shift+P expands or collapses the panel, and any check you find noisy can be switched off in settings.

PRIVACY: Prompt Check runs entirely in your browser. There is no account, no API key and no server. Your prompt text is analysed locally and is never transmitted anywhere. The extension makes no network requests of any kind. This is verified automatically on every build.