Hermes Agent's hazard command approval should be on by default. The official Security documentation divides the approval modes into three categories: manual, smart, and yolo. For most users, manual is the most stable, smart is suitable for slightly more mature automation environments, and yolo is only suitable for sandboxes where you can fully tolerate the consequences.
How to choose the three modes
- manual: You need to manually approve every high-risk operation, which is the safest.
- smart: Low-risk commands are automatically released, and high-risk commands are still required to be confirmed, suitable for frequent daily operations.
- yolo: Basically unstoppable, fast but the highest risk.
How will it behave on the chat platform?
On messaging platforms like Telegram, Slack, and Discord, Hermes will send the command details to the chat when encountering a dangerous command, waiting for you to approve it clearly. In other words, the approval system is not unique to the CLI, but is also important in remote gateways.
Advice for newbies
Run a week with manual to figure out what commands Hermes usually triggers; After confirming that your tool scope, directory permissions, and user permissions are reasonable, consider switching to smart. If you're still figuring out the structure of the directory, don't use /yolo for a long time.
In a word: Hermes Agent's most worthy "slow step" is dangerous order approval; It will save you more time on rework.
Official open source address: https://github.com/NousResearch/hermes-agent; Official document entry: https://hermes-agent.nousresearch.com/.
