Triggers

Calendar triggers start work from a scheduled agent prompt.

This is the clearest trigger in CLEA today. The user chooses a schedule, writes the prompt, and can mention workflow files that CLEA should follow.

Manual chat triggers start when the user asks CLEA to run a workflow.

This is the safest way to test a workflow. Run it once, inspect the result, then decide if it should become scheduled work.

Prompt-run triggers are follow-up routines after visibility prompts create fresh answer data.

For example, a daily visibility run can be followed by a CLEA prompt that reviews new chatlogs, sources, and competitor mentions.

Score or data-change triggers should be handled carefully.

In practice, this means asking CLEA to check whether a metric changed enough to deserve action. Do not assume every small move should create a task.

A trigger should pass enough context for CLEA to work.

That context can be the schedule, the workflow file, dashboard data, prompt results, source data, or a page file.

Noise is the main danger.

If every small change creates a task, users stop trusting the workflow. Add clear rules for when CLEA should act and when it should return `no action`.