Define the routine

The first step is to decide what CLEA should repeat.

Keep the routine close to a real dashboard decision. Good routines start from questions like:

• Which prompt got weaker?
• Which source pattern repeated?
• Which competitor gained mentions?
• Which page should we improve first?
• Did the new page change the answer?

Write the routine as a small file.

Use simple headings. CLEA does not need fancy formatting. It needs clear instructions.

Goal: review weak visibility prompts.
Inputs: Analytics, related prompt answers, Sources, Competitors.
Steps:
1. Find the prompt with the clearest drop.
2. Read the answer text.
3. Check sources and competitor mentions.
4. Decide if this is a prompt issue, source issue, or page issue.
Output: one short table and one suggested next action.

The output should match the routine.

If the routine is research, ask for a table. If it is action, ask for one suggestion card. If it is page work, ask for a page brief.

Run the routine once from chat before scheduling it.

If CLEA gives a broad answer, narrow the routine. If CLEA misses evidence, name the exact dashboard view or file it should inspect.