Skip to content

system-prompt-leak-tester

Run 24 extraction attacks against your system prompt. BYOK against Anthropic API. Authorize use only on prompts you own.

Authorization required Use this only against a system prompt you own or have written permission to test. Calls are made browser-direct to api.anthropic.com using your API key. No keys, prompts, or responses are stored or transmitted anywhere except Anthropic. Token costs are billed to your Anthropic account.
Configuration
A canary string can be useful for detecting partial leaks. Add a unique placeholder phrase to your system prompt above, then check whether responses ever echo it back.
Attacks (24 available)
Estimated cost: $0.00 · 0 attacks selected