Skip to content

Security: meSingh/polinrider-cleaner

SECURITY.md

Security policy

Reporting a vulnerability in these scripts

If you find a way to make a script in this repository destroy data, restore a branch to an attacker-controlled commit, leak a token, or execute attacker- controlled input, report it privately through GitHub's Report a vulnerability button on the Security tab. Do not open a public issue.

Include the script, the exact command, and what an attacker gains.

What is in scope

  • Any path where a script deletes data, or changes state without --apply
  • A restore that can be steered onto a malicious commit
  • Token or credential exposure: in output, in a report file, in a clone URL, or in a process argument list
  • Command injection through a repository name, branch name, file path or indicator file

What is not in scope

Trusting this repository before you run it

You are being asked to run shell scripts during a security incident, which is exactly the situation where you should be suspicious.

  • Every script is plain shell or PowerShell. Read it. The longest is just over 400 lines.
  • Nothing fetches code at runtime. The only network calls are git clone, gh api, and the checkout action in CI.
  • Everything except restore.sh --apply and the local checks with --apply is read-only.
  • Pin what you run: clone at a commit SHA you have read, not a moving branch.

There aren't any published security advisories