Prompt Security Eval Playground
prompt security eval·Medium
time to bug
00:00
median 06:30
Analytics cookies
We use PostHog to collect page views, feature usage, and anonymous error reports. No email or passwords are sent. Privacy policy
Practice spotting prompt-injection and agent-harness discipline failures in a mocked evaluator.
No output yet
Confidence: —
Run the harness to inspect the system prompt.
No trace available