Skip to content

Security: CedricPoint/cmdguard

Security

SECURITY.md

Security policy

What cmdguard protects against

Accidents and careless generation: a recursive delete rooted in an empty variable, a force push to main, a curl … | sh in a README someone pasted.

What it does not protect against

An adversary who already runs code on the machine. cmdguard inspects a command string; it does not sandbox, does not resolve variables, and does not follow the contents of scripts on disk. Anything that hides the real command — base64, a string assembled at runtime, . ./script.sh — will get past it. Treat it as one cheap layer, not as the boundary.

Reporting a bypass

A bypass is a command that a reasonable person would call obviously destructive and that cmdguard returns allow for.

Open a regular issue with the command line — there is nothing confidential in a missing rule, and a public issue gets it fixed faster. If you believe you have found something that should not be public (for example an arbitrary-code-execution bug in cmdguard itself, not a missing rule), use GitHub's private vulnerability reporting instead.

Supported: the latest release on the main branch.

There aren't any published security advisories