Paste a prompt template and get a report: unfilled variables, an approximate token and length estimate, potential prompt-injection patterns, and simple quality checks. All heuristics run locally.
Token count is a rough estimate (about 4 characters per token). Real tokenizers differ by model.