Skill Template: Trending Knowledge Agent Topics

The weekly automation described in Maintaining your Knowledge Agent over time is most useful when it's configured to surface exactly what your team needs. Rather than building a trend digest skill from scratch, you can use the prompt below to have Operator Mode build it with you — asking the right questions and drafting the skill based on your answers.


Before you start: connect your delivery tool

If you want the digest delivered to Slack or another external tool, that tool needs to be connected to Guru as an MCP integration before the skill can send anything to it. Without an active connection, the digest will run but the output will only be visible inside the Agent Center.

If you're using Slack: Make sure the Slack MCP is connected and authenticated in your Guru workspace. See Connecting MCP integrations for how to set this up.

If you're using a different tool: The same applies — any external delivery destination requires an authenticated MCP connection for that tool.

If you're not sure whether your integration is connected, check under Apps and Integrations in Guru before moving to Step 1.


Step 1: Open Operator Mode and paste this prompt

Open Operator Mode for your Knowledge Agent. Paste the following prompt and send it — Operator Mode will ask you each question one at a time, then draft the skill based on your answers for you to review before anything is saved.

I want to build a skill that creates a recurring trend digest for a Knowledge Agent — 
a regular summary of where the agent is struggling and where the knowledge base has 
gaps. I don't want to build this from scratch. Instead, please ask me the following 
questions one at a time, wait for my answer before moving to the next, and then use 
my answers to draft the skill for me to review before saving.

Here are the questions to ask me:

1. Which Knowledge Agent should this digest cover? If I manage more than one, 
   I may want this scoped to a specific agent.

2. How far back should the digest look each time it runs — 7 days, 14 days, 
   or a different window?

3. How many trends should it surface per digest? Fewer trends (3–5) keep the 
   output focused. More trends give broader coverage but take longer to review.

4. What kind of gaps matter most to me? For example: questions that went 
   completely unanswered, questions that got wrong or unhelpful answers, 
   questions on a specific topic area I'm trying to improve, or all of the above.

5. Who should receive the digest when it runs — a Slack channel (if so, which 
   one?), a direct message to me, or kept inside the Agent Center only?

Once you have my answers to all five questions, draft the skill. The skill should 
include clear instructions for what to look for, what to exclude, how to format 
the output, and how to deliver it. It should also include a note that this skill 
surfaces information for the owner to act on — it does not create cards, change 
agent configuration, or resolve gaps automatically.

Before saving, show me the full draft and ask me to confirm.

Operator Mode will work through the questions with you and produce a skill draft tailored to your answers. Review it carefully before approving — especially the delivery method, which requires an active MCP connection if you've chosen an external tool.


Step 2: Attach the skill to an automation

Once the skill is saved, set up a Knowledge Agent automation to run it on a recurring schedule. A weekly cadence works well for most teams — frequent enough to catch gaps before they accumulate, without producing more output than you have time to review.

💡

Tip

If the digest is surfacing more trends than feel manageable after a few weeks, go back into Operator Mode and ask it to revise the skill with a lower trend count. Three to five well-described trends you can act on is more useful than ten that sit unread.




Did this page help you?