Skip to content

Security: PeteAndrews1289/pokemon-red-ai

SECURITY.md

Security Policy

Supported version

This repository is a completed research record. Security and reproducibility fixes are supported on main; historical experiment branches, tags, and external run artifacts are preserved as evidence and are not maintained as active releases.

Reporting a vulnerability

Please use GitHub's private vulnerability reporting for vulnerabilities involving the harness, artifact validation, checkpoint handling, command-line interfaces, or CI.

Do not include a ROM, save file, emulator snapshot, model checkpoint, credential, private machine path, or unredacted runtime trace in a report. Describe the issue with the smallest synthetic fixture that reproduces it.

Useful reports include:

  • a path that bypasses the private-artifact publication guard;
  • unsafe loading or deserialization of an untrusted checkpoint or run record;
  • command or path injection through a CLI argument;
  • an integrity check that accepts a modified ROM, checkpoint, or evidence file;
  • a boundary violation that lets trainer-only state choose or replace an evaluated action; or
  • a CI or documentation path that publishes proprietary or sensitive artifacts.

Research boundary

Hashes in this project provide identity and integrity checks; they are not digital signatures and do not authenticate an author. The emulator and learning dependencies process user-supplied local files, so only lawfully obtained, trusted game and checkpoint files should be used.

There aren't any published security advisories