The reason why commands such as /status and /disk in Hermes Agent are returned in seconds is not because the model suddenly becomes faster, but because they do not follow the model at all. The official configuration document clearly defines Quick Commands: they execute local shell commands directly, do not call LLMs, have zero tokens, and return faster.
What is this kind of command suitable for?
- Check the service status
- Check disk space
- Run fixed O&M scripts
- Do low-risk, fixed-output quick checks
It is different from ordinary dialogue
The general dialogue is "the model thinks and decides what to do"; Quick Commands is "execute directly when the command name is matched". So it's fast, but also more rigid, suitable for fixed actions, not for complex reasoning.
What else to pay attention to
The official documentation states that there is a 30-second timeout, and long commands are killed; And Quick Commands take precedence over skill commands of the same name, so don't collide with your usual skills when naming them.
In a word: /status is not an optimization miracle, but because Hermes Agent directly ran the local command.
Official open source address: https://github.com/NousResearch/hermes-agent; Official document entry: https://hermes-agent.nousresearch.com/.
