Skip to content

Security: Bensonluo/rag_chatbot

Security

SECURITY.md

Security Policy

Scope

This repository runs a public demo (benluo.art/projects/rag-chatbot) of an agentic-commerce and RAG customer-service engine. The commerce catalog is a deterministic synthetic dataset and payments are simulated — no real money moves.

Security issues in the following areas are still in scope and taken seriously:

  • Authentication, session, and guest-cookie handling
  • Authorization boundaries (session ownership, admin endpoints, knowledge-base management)
  • Prompt injection and guardrail bypasses that could trigger unintended business actions
  • The commerce state machine (idempotency, exactly-once semantics, confirmation gates)
  • Injection vectors in retrieval, ingestion, or template rendering
  • Anything in a dependency that is exploitable in the demo deployment

Reporting a Vulnerability

Please do not open a public issue for security reports.

Use GitHub's private vulnerability reporting: Security tab → Report a vulnerability on this repository. If private reporting is unavailable, open a regular issue asking the maintainer to enable it, without vulnerability details.

Include when possible:

  1. Affected component or endpoint (e.g. app/api/v1/..., /commerce/*)
  2. Steps to reproduce, with the exact request/payload
  3. Impact assessment (what an attacker gains)
  4. Suggested fix, if any

Response Expectations

This is a portfolio/open-source project maintained in spare time. Aim for:

  • Acknowledgment within a few days
  • Assessment and (if accepted) a fix in a reasonable timeframe
  • Credit in the release notes if you wish

Demo Boundaries

The live deployment intentionally keeps secrets out of the repository, uses token-gated internal services, and runs simulated commerce only. Reports that require physical access, compromised credentials, or control of the hosting server are out of scope.

There aren't any published security advisories