What Scheduled tasks is for
Hermes documents one-shot and recurring tasks, skill attachments, management actions, and output delivery to configured destinations. Scheduled agent sessions can be useful for briefings, monitoring, and maintenance. A schedule should follow a successful manual example with explicit success criteria. For deterministic checks, the docs also describe script-only tasks without model involvement; use an agent when judgment or synthesis is actually needed.
What you need
- A running scheduling/gateway environment and the credentials needed during unattended execution.
- A clear timezone, source window, output location, delivery destination, and acceptable no-change behavior.
Set it up
- 1
Run the task manually on a small input and inspect the result. Remove steps that depend on an interactive prompt.
- 2
Create the schedule through the documented management interface, attach only necessary skills, and define bounded retries plus a concrete failure report.
Check it works
- Observe a real scheduled attempt and read its saved output. Verify that freshness and required sections match expectations.
- Check the intended destination independently and confirm that duplicate runs do not repeat irreversible actions.
Limits and failure points
- An enabled job does not prove the host was awake, sources were reachable, or delivery succeeded.
- Model selection and credentials can affect later runs. Review effective configuration and preserve the last good output when a task fails.
Primary sources
Spotted something wrong or out of date? Tell us on X at @HermesWatcher (opens in a new tab) and we will fix it.