Skip to content

Security: csd113/Vitrallis-Apps

Security

SECURITY.md

Security policy

Security reports are welcome for Vitrallis Apps packages, catalog metadata, validation, and publication tooling. Installer, launcher, and App Center issues belong to Vitrallis Shell; consult that project's security guidance for its reporting process.

Reporting a vulnerability

Do not put exploit details, malicious payloads, access codes, API keys, or private device data in a public issue or pull request.

Use Report a vulnerability on the repository's Security page when available. Private vulnerability reporting is currently disabled, and no dedicated private security address is published. Until the owner provides a private channel, open an issue titled “Private security contact requested” with no technical or identifying details, or ask the repository owner through a contact method they publish on their GitHub profile. Wait for a private channel before sharing the report. A public issue is not a confidential channel.

Once a private route is established, include:

  • Affected app/tool, version, catalog source pin, and Shell version if relevant.
  • Device/OS/runtime, the trust boundary crossed, and the expected impact.
  • Minimal reproduction steps using synthetic files and disposable storage.
  • Any mitigation you have verified, without including real credentials or user data.

There is no published response-time or long-term support commitment. Vitrallis is pre-release; start from the current catalog version and report older affected versions too. Do not assume an old version receives backported fixes.

Trust and deployment boundaries

  • Catalog pins and SHA-256 checksums let clients verify advertised bytes. They do not independently authenticate the publisher or make untrusted code safe.
  • Manifest network/audio/storage flags declare requirements. Apps run with the launching user's privileges and are not sandboxed by this format.
  • Validate external metadata, files, paths, and source inventories before use. Keep credentials, caches, and private data out of packages: all app files outside tests/ are published.
  • Media Carousel's management interface uses unencrypted HTTP and a per-launch access code. Use a trusted LAN and do not forward its port to the Internet. Follow its security and limits, and keep Pillow and FFmpeg current.

For the exact validation and client responsibilities, see catalog trust boundaries and runtime integration. Ordinary setup problems and non-sensitive bug reports go through Support.

There aren't any published security advisories