Prompt Analyzer

A second look at your instructions, with an explanation for every suggestion.

Structural analysis

The Analyzer checks for empty text, required placeholders, recognizable task markers, output format, structured context, a short list of vague phrases, known literal format conflicts, repeated instruction lines, unclosed delimiters, long instructions, and an image subject where applicable.

What a finding means

“Return JSON only” and “Write in Markdown” trigger a known-format conflict. That is an invitation to review your intended format. It is not a general contradiction detector. Recognized code fences, quoted lines, and source_material blocks are excluded from instruction phrase checks.

What it cannot tell you

The tool does not verify facts, understand whether context is sufficient, predict hallucinations, prove safety, or guarantee effectiveness. Free-text context and output format may be marked Not assessed. Concise prompts do not need extra roles or adjectives to earn approval.

Repairs are your choice

Delimiter repairs have a before-and-after preview and an Undo checkpoint. Findings belong to the exact document revision; editing or changing the document makes the report stale until you run it again.

Counting behavior

Characters are Unicode code points. Words are whitespace-separated groups. Lines are counted from line breaks; an empty editor has zero lines. There is no model-independent token count.