QuivrHQ vulnerabilities

Quivr versions through 0.0.322 fail to validate chat ownership in the GET /chat/{chat_id}/history, DELETE /chat/{chat_id}, and POST /chat/{chat_id}/question/answer endpoints. Authenticated attackers can read other users' conversation histories including private knowledge base content, delete arbitrary chats, and inject fabricated messages into other users' conversations.

Quivr through 0.0.322 fails to validate ownership in prompt endpoints, allowing authenticated users to modify any prompt by identifier. Attackers with read-only access to shared brains can read exposed prompt identifiers and overwrite system prompts affecting all brain users.