Agent prompts

Agent prompts tell CLEA to do work with the evidence already in the workspace. They can inspect prompt results, review sources, summarize patterns, create suggestion cards, or prepare a brief.

They are not the same as mention prompts. Mention prompts measure visibility. Agent prompts act on evidence.

Use mention prompts when you want to know what AI systems answer.

Use agent prompts when you want CLEA to interpret results or create an output from them.

Agent prompts get better when they reference the right files and dashboard context.

Use files for repeatable routines. Use dashboard context when the agent needs current prompt results, sources, competitors, or suggestions.

Tell the agent what to return. This prevents broad summaries that are hard to use.

Good prompts ask for one suggestion card, one team summary, one page brief, or one short priority list.

Run a prompt manually before scheduling it. This gives you a chance to check the evidence, output shape, and review burden.

Once the result is useful, schedule it on the calendar.

Review agent results against the evidence. The agent should be able to explain which prompt runs, sources, files, or analytics views shaped the output.

Do not accept public claims or page changes without review.

Reusable prompt examples make CLEA easier to operate. Keep them simple, concrete, and tied to an output.

Goal: inspect prompt results and create one useful next step.
Inputs: latest chatlogs, mention outcomes, source analytics, competitor movement.
Rules: ignore one-off noise, cite the evidence, do not publish.
Output: one suggestion card with a clear CTA.

Review the latest prompt run for this buyer question.
Tell me if our brand was named, who appeared instead, and which source seems most important.
End with one next step.