askdare exposes a public, anonymous chat endpoint that spends money on your OpenAI account. We take vulnerabilities in it seriously.

During 0.x only the latest minor release receives security fixes.

Please do not open a public issue. Report vulnerabilities privately through GitHub's private vulnerability reporting: open the Security tab of this repository, then Report a vulnerability.

If you can't use GitHub, email hello@boldare.com with the subject "askdare: security". Keep the details for the private channel we will set up with you.

Please include:

• the affected version and configuration;
• steps to reproduce, or a proof of concept;
• the impact you expect (for example, cost abuse, data exposure, XSS).

We aim to acknowledge reports within 3 working days and to agree on a disclosure timeline with you.

In scope:

• authentication and session tokens;
• conversation ownership;
• rate limits and spend budgets;
• the embed handshake (postMessage, frame-ancestors);
• rendering of model output (XSS);
• upload validation;
• leakage of server configuration to the client, including what getChatConfig returns;
• rating handles and POST /feedback;
• the chat inside a page (@boldare/askdare/react): its same-origin checks and output rendering;
• the add-ons @boldare/askdare-upstash and @boldare/askdare-otel.

Prompt injection that causes only off-topic answers is a known limitation of LLMs and is out of scope. Prompt injection that triggers tool misuse, data exposure or cost amplification is in scope.

The threat model is in THREAT_MODEL.md; complete the going-public checklist before any public deployment.