Hermes Agent follows up on Claude Code's/loop: Agents can wake up and work on their own at regular times

Hermes Agent follows up on Claude Code's/loop function. Users can ask the Agent to perform the same task repeatedly in the current session, such as watching CI, checking deployment status, or repeatedly repairing the code until the test passes. Each trigger is a complete Agent execution, and the latest state at that time is re-read. Users can specify how often it should be executed, or they can hand over the frequency to Hermes. Adaptive mode defaults to checking every minute. If the results do not change for several consecutive times, it will be reduced to 2, 4, and 8 minutes. Once the results change, it will be restored to once every minute for a maximum of 15 minutes. The number of cycles and stopping conditions can also be set. The Agent can judge that the task has been completed by himself, or use/loop --until to have additional judges check whether the conditions are met. By default, it will pause after running 100 times to avoid burning tokens when unattended. This feature directly follows the/loop of Claude Code and is even compatible with the/proactive alias.