AnyoneLearnAI
Start learning

Lab · Security · ~3 min

Prompt-injection sandbox

Paste a user message and see which phrases look like prompt injection. Educational pattern check — not an attack tool.

2 injection-shaped patternsThis is a teaching regex, not a production filter. Real attacks hide in documents, tools, and encodings. Treat untrusted text as data, not instructions.

  • Ignore / override previous instructions
  • Reveal system / hidden prompt

How to read this

  • Paste text that would arrive from a user, a retrieved document, or a webpage.
  • A hit means “looks like a known pattern,” not “this is a working exploit.”
  • Continue in the prompt-injection hub for defenses that belong in the system.
Continue in Prompt injection →All labs