Scheduled agent prompts

Scheduled agent prompts tell CLEA to repeat a task on a schedule. They are useful after you know the task produces value.

Do not schedule a vague instruction. Schedule a routine that already works.

The schedule should match the signal. Daily visibility checks can be useful, but deeper reviews often work better weekly.

Use the calendar to see what runs when and avoid stacking too many tasks on the same day.

Reference workflow files when the routine needs detailed instructions. This keeps the scheduled prompt readable.

The file should explain inputs, review steps, output shape, and approval boundaries.

Define the output in the prompt. A good scheduled agent might create one suggestion card, one weekly summary, or one page writer brief.

Clear output makes review easier.

Strong scheduled prompts are specific, evidence-based, and limited.

Every Monday, review the latest visibility prompt runs.
Use source analytics and competitor movement.
Create one suggestion card only if there is a clear action.
Do not suggest publishing changes without human review.

The first scheduled run is a test. Check whether CLEA used the right evidence, ignored noise, and produced the expected output.

If the first run is weak, tune the prompt before repeating it.

Scaling a bad schedule creates a noisy workspace. Pause or tune routines that create vague cards, duplicate reports, or unsupported claims.

Automation should make the workspace calmer, not busier.