Suggestion-card output

Suggestion-card output agent behavior

Suggestion-card output is the right format when the agent finds one next action.

A card should not summarize everything. It should help the user decide what to do next.

Configuration point

Tell the agent when a card is allowed. For example, only create a card when the evidence is clear, the action is specific, and the CTA points to the right surface.

This prevents noisy suggestion queues.

Review source gap on comparison prompts

Competitors are named from review pages where your brand is missing. Inspect the prompt evidence and decide whether this needs a page update or source review.

Open source analyticsAsk CLEAArchive

Context to include

Include the evidence that supports the card. That can be a prompt result, source pattern, competitor movement, readiness issue, or workflow file.

If the agent cannot point to evidence, it should not create a strong recommendation.

1

Evidence

Name the prompt, source, competitor, readiness item, or file.

2

Issue

Explain the problem in plain words.

3

CTA

Point to the next smallest useful action.

4

Review

Let the user accept, ask, complete, or archive.

Output review

Review the card title, reasoning, evidence, and CTA.

Archive cards that are vague, duplicated, or not tied to a real workspace action.

Next step

Main website