Dynamic samples using actual prompt-guard patterns.
Supported attack types: 46
Supported languages: en, ko, ja, zh, es, de, fr, ru, pt, vi
| Attack Type | URL |
|---|---|
| instruction_override | /test/test?type=instruction_override |
| role_manipulation | /test/test?type=role_manipulation |
| system_impersonation | /test/test?type=system_impersonation |
| output_manipulation | /test/test?type=output_manipulation |
| data_exfiltration | /test/test?type=data_exfiltration |
| approval_expansion | /test/test?type=approval_expansion |
| dm_social_engineering | /test/test?type=dm_social_engineering |
| emotional_manipulation | /test/test?type=emotional_manipulation |
| authority_recon | /test/test?type=authority_recon |
| phishing_social_eng | /test/test?type=phishing_social_eng |
| urgency_manipulation | /test/test?type=urgency_manipulation |
| cognitive_manipulation | /test/test?type=cognitive_manipulation |
filetype: html, md, txt, json, xmlseverity: critical, high, medium, lowtype: 46+ attack typesobfuscated: true, falselang: en, ko, ja, zh, es, de, fr, ru, pt, viclean: true (returns clean content)