LLM Prompt Injection Tester

Heuristic-scan your system prompt for injection weak spots

Tests pasted content against a built-in library of ~18 known prompt-injection attack patterns — ignore-instruction phrasing, DAN/jailbreak personas, fake delimiter injection, and more — while separately running a checklist-based hardening assessment of your system prompt itself. Produces a 0-100 score with concrete suggestions for tightening instruction boundaries. Clearly flagged as an educational, heuristic tool, not a substitute for real red-teaming against your target model.

• →Scans sample user input against ~18 known prompt-injection phrasings, including DAN-style jailbreaks and fake ### SYSTEM ### delimiter injection
• →Runs a 5-point checklist against your system prompt: clear boundaries, override resistance, non-disclosure, scope, and detail
• →Produces a 0-100 hardening score with specific, actionable suggestions for each failed check
• →Explicitly labeled as an educational heuristic checker — not a guarantee of security against real attacks
• →Watermarked by Cikal Studio Labs · Works on any device, no install required