IBM’s big pitch today is a separate identity for every agent. You don’t need an enterprise platform to get most of the benefit at home. Here’s the pattern I’d use for any coding or blog agent that touches GitHub or a site’s API.
1. One agent, one credential
Never paste your own all-access token into an agent. Make a new one per bot. On GitHub, use a fine-grained personal access token limited to the one repo the bot works in, with only the permissions it needs (often just Contents and Pull requests), and a short expiry.
2. Store it where only that bot can read it
Put it in an env file owned by the user the agent runs as, then chmod 600 it. Keep it out of the repo and add the file name to .gitignore. If your agent runs in a container, mount it as a secret instead of baking it into the image.
3. Least privilege on the other side too
For WordPress, a bot that writes drafts gets its own account and its own Application Password, so you can revoke just that one without touching yours.
4. Rotate and revoke
Put the expiry date on your calendar. If a bot ever acts weird, revoke its token first and ask questions later. The LiteLLM supply-chain mess started with one exposed master key. Don’t hand anything a master key.
Review before you run anything: read the docs yourself, back up your config, and try it somewhere you can afford to break first.
God Bless, and Happy hacking!
