$ open_llm_demo --no-guardrail

prompt: (none — try ?prompt=hello)
model:  llama-3.1-8b-instruct (Workers AI)
output: (no prompt provided — POST or ?prompt= to test)

This endpoint is a known security-research honeypot operated by TOUGH LOVE SECURITY. Captures are used to harden the AgentShield Constitutional classifier (see /constitutional). Source-of-truth: /AGENTS.md.