Puzzle Lab / AI traps
Model traps and prompts
Puzzles that teach LLM boundaries, tokenization, tool use, and prompt-injection thinking.
Showing 2 of 2 objects in this shelf.
- Prompt Injection Museum Guard
An AI reads a web page. The page tries to trick it. What should the AI ignore?
- Tokenizer Trap
LLMs read chunks of text called tokens. That makes some simple-looking counting tasks harder.
No matching item on this page. Clear filters or use the Explore filter for the full archive.