colabctl ships frequently; only the latest released minor gets security fixes. Please upgrade with colabctl update before reporting.

Report security issues privately through GitHub's private vulnerability reporting — open the repository's Security tab → Report a vulnerability (direct link). Please do not open a public issue for a vulnerability.

Include the affected version, a description, reproduction steps, and the impact. We aim to acknowledge within a few days and to coordinate a fix and disclosure with you.

colabctl handles credentials and drives remote runtimes, so the security-relevant surface is:

• credential handling — the OS-keychain / encrypted-file secret store, environment injection, and proxy-token caching;
• the custom native /tun/m/* transport and the detached-job runner;
• code/argument marshalling for @remote and detached jobs (cloudpickle by value).

Out of scope (report upstream): Google Colab itself, third-party backends (Modal, Vertex, RunPod, Vast, …), and the tracking libraries (Weights & Biases, MLflow).

By design, credentials are never written into the state document. Secrets live in the OS keychain or an encrypted file; experiment-tracking credentials are injected as environment variables (never baked into pickled code) and re-resolved from the secret store on every run rather than persisted. If you find any path where a credential is logged or persisted in plaintext, that's a vulnerability — please report it.