PromptLint is a static analyzer for LLM prompts — think ESLint, but for the text driving your AI pipelines.
Most teams ship prompts with zero guardrails. PromptLint catches the bugs before they hit production: injection attacks, token waste, vague language, and missing structure. No API calls, no latency, results in milliseconds.
What it catches:
- Prompt injection: detects role-hijacking and "ignore previous instructions" patterns before they reach your pipeline
- Token bloat: flags politeness filler, redundant phrasing, and structural waste burning your API budget
- Missing structure: warns on prompts lacking delimiters, output format specs, or role definitions that degrade output quality
- Cost projection: estimates token cost at real-world scale so you know the budget impact before you ship
How it works:
- VS Code extension: inline linting as you type, squiggly underlines and lightbulb auto-fix, just like ESLint
- CLI tool: drop it into any pipeline as a pre-commit hook or GitHub Action, fails builds on CRITICAL issues
- Auto-fix: safe changes applied automatically, ambiguous rewrites shown as suggestions you approve
Built for prompt engineers, AI developers, and any team shipping LLM features who wants their prompts to be secure, efficient, and consistent.