Skip to content

Security: Samurai33/CyberJust

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Please don't open a public GitHub issue for a security vulnerability.

Use GitHub's private vulnerability reporting for this repository instead — it notifies the maintainer directly without exposing the report publicly until a fix is ready.

Include what you can:

  • What the vulnerability is and where (file/line if you have it)
  • Steps to reproduce
  • What an attacker could actually achieve with it

There's no dedicated security team or SLA — this is a small project maintained by one person. Response time is best-effort, not guaranteed within a fixed window.

Current state

CyberJustiça Brasil is a fully static site (Astro build served by nginx-unprivileged). There is no backend, no database, no login, no forms and no user data on the server.

  • Headers (docker/nginx.conf, docker/security-headers*.conf): CSP with no 'unsafe-inline' and only 'self' sources, frame-ancestors 'none', HSTS, COOP, CORP, nosniff, Referrer-Policy and Permissions-Policy. The e2e suite fails on any CSP violation.
  • Browser storage: the audio player keeps playback position, speed and volume in localStorage. It never leaves the device.
  • Third parties: none. Fonts, images and audio are all served from this origin. There is no analytics and there are no cookies.
  • Supply chain: exact dependency versions and a frozen lockfile. pnpm minimumReleaseAge only installs releases that are 14+ days old, and only esbuild/sharp may run install scripts. Docker base images are pinned by digest and GitHub Actions by commit SHA. Dependabot runs with a 14-day cooldown.
  • CI: pnpm test + astro check + build, and an e2e run against the production Docker image.

This section will get out of date. If it no longer matches the code, that mismatch is itself worth reporting.

Scope

In scope: this repository's code (src/, docker/, Dockerfile, .github/workflows/).

Out of scope: third-party services this project depends on (report those upstream), social engineering, physical access, and denial-of-service reports without a concrete exploitable flaw.

There aren't any published security advisories