Skip to content

Security: OpenAgentis/chatdify

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Please do not open a public issue. Email tfanac@cse.ust.hk with a description, steps to reproduce and the affected version.

Things to know when running ChatDify

  • Credentials. LLM keys and Dify passwords are read from ~/.chatdify/config.toml (created with mode 600), environment variables or .env. They are never written to generated DSL. Logs in ~/.chatdify/logs/ contain prompts and LLM responses; treat them as sensitive.
  • Generated code is untrusted. Workflows may contain code nodes and HTTP requests written by an LLM. Review them before importing into a production workspace.
  • Dynamic validation executes workflows. With --dify-url, ChatDify imports and runs generated apps, including their code nodes (inside Dify's sandbox) and outbound HTTP calls. Use a dedicated, non-production Dify workspace and account.
  • Templates. The knowledge base in chatdify/kg/ contains placeholders such as YOUR_API_KEY. If you find a real credential in the repository, report it as above.

There aren't any published security advisories