Skip to content

Security: eseeyuh/courseflow

SECURITY.md

Security Policy

Supported versions

CourseFlow is pre-release (v0.1.0 in development). Only the latest main branch is supported.

Reporting a vulnerability

Please do not open a public issue for security problems.

Report privately through GitHub's private vulnerability reporting for this repository (Security tab → "Report a vulnerability"). Include steps to reproduce, the affected component and the potential impact. You should receive an acknowledgement within a few days.

Security model (summary)

  • Course content is untrusted. Text in uploaded or imported documents is treated as data, never as instructions. Prompt-injection content must not change workflow control flow or tool use.
  • Allowlisted, typed tools only. The AI workflow has no shell access, no arbitrary SQL and no open internet access. Tool arguments are validated server-side.
  • Secrets stay out of prompts and git. Credentials live in environment variables only. .env files are git-ignored, and only .env.example with placeholders is committed.
  • Read-only towards the LMS. CourseFlow never submits, edits or grades university content.
  • Synthetic demo data. The repository contains no real university material or student data.

A dedicated threat model will be added as the system is implemented.

There aren't any published security advisories