Scheduled agent prompts

Scheduled agent prompts give CLEA a recurring job. They are useful when the same review should happen every day, week, or reporting cycle.

Schedule a routine only after the manual version has produced a useful result.

The calendar shows when the agent prompt runs. Use it to keep visibility checks, source reviews, reports, and content routines from overlapping too much.

Cadence should match the value of the signal.

The prompt should explain the task, evidence, decision rule, and output.

Avoid instructions like "review everything". They create vague output.

Referenced workflow files keep scheduled prompts short. The file can carry the detailed steps and review rules.

This is useful when a routine becomes part of the workspace rhythm.

Every Wednesday, inspect source movement from the latest prompt runs.
Look for repeated sources that support competitors.
Create one suggestion card only when the action is clear.
Skip the card if the evidence is weak or repeated from last week.

The best scheduled outputs are small and reviewable. Examples include one suggestion card, one weekly summary, one source gap, or one prompt cleanup recommendation.

Do not schedule outputs that nobody will read or complete.

Review the first few scheduled runs carefully. Look for duplicate cards, weak reasoning, missing evidence, or tasks that arrive too often.

Tune the prompt before increasing the schedule.

Noisy routines reduce trust. Pause them when they create generic cards, repeat old tasks, or ignore the current workspace state.

Automation should make the workspace clearer.