Prompt Linter
Run local static checks on your prompt: unresolved placeholders, over-long text, missing system messages, suspicious JSON, and redundant wording — no LLM calls.
All processing for core tools runs locally in your browser. Privacy policy
0 characters0 / 128,000 tokens1 errors0 warnings0 tips
Prompt under test
Findings
- ErrorEmpty prompt
There is no content to lint. Paste a prompt or load from the Editor.
How to use
- Load a prompt from the Editor, or paste/edit in the left pane.
- Choose Text or ChatML and the target model for context checks.
- Review errors, warnings, and tips on the right — all rules are local and static.
- Fix issues here or open in the Editor / Cleaner / Variable Batch as needed.
FAQ
Does the linter call an LLM?
No. Every rule is a deterministic local check. There is no AI rewrite and nothing is uploaded.
What counts as an unresolved placeholder?
Double-brace names like query or lang that still appear in the prompt. Use Variable Batch to fill them before shipping.
Are token counts exact?
They use the same local tokenizer stack as the Editor (Exact for OpenAI encodings; Estimate for some other vendors).