This project is pre-1.0. Security fixes target the main branch until the first
tagged release plan is published.
Do not open a public issue for suspected vulnerabilities.
Email security reports to research@worldflowai.com with:
- affected version or commit;
- reproduction steps;
- impact;
- whether the issue can expose prompts, token IDs, request metadata, or KV data;
- any suggested mitigation.
We will acknowledge credible reports within 5 business days when possible.
This connector should avoid retaining prompt text by default. Keep
enable_prompt_text=false unless a deployment explicitly opts in and has the
right data-handling controls.