Understanding Rovo Agent limitations: auto-creating vs. publishing Confluence pages

[Okan Erdogan]

Hi everyone 👋

I’m currently building a Rovo Agent to support a Managed Services documentation workflow, and I’d love to sanity-check my understanding of Rovo’s current capabilities vs. limitations.

High-level use case

• We receive Jira Service Management (JSM) tickets from clients.

• For every ticket, we create a standardized Confluence documentation page:

  • Uses a predefined template

  • Is placed under a specific parent page (e.g. Managed Services → Support Tickets)

  • Pulls context from:

    • the JSM ticket itself

    • a project-specific “Managed Services – Project Delivery Configuration” page in Confluence (deployment model, env flow, testing expectations, etc.)

• The agent can successfully:

  • resolve the correct project

  • locate the correct Confluence space via configuration pages

  • generate accurate, structured content

  • respect strict knowledge-scoping and non-hallucination rules

What I’m observing

Even when:

• the JSM project is linked to its Confluence space

• the agent can deterministically identify the correct space and parent page

• permissions are in place

…the agent still requires a manual space selection step at publish time (i.e. “Select a space to publish the page”), and does not fully auto-create/publish the page without human confirmation.

My question

Is this an intentional platform limitation of Rovo today?

Specifically:

• Can Rovo Agents currently prepare Confluence pages but not autonomously publish them to a resolved space (even when that space is known and accessible)?

• Is the manual space-selection step expected behavior for governance/safety reasons?

• Are there any roadmap plans (or best-practice patterns) for trusted auto-publish in the future?

The workflow still provides huge value (80–90% time savings), so this is more about understanding where the hard boundary is today, not a complaint.

Thanks in advance — really enjoying what’s possible with Rovo so far :)

P.S. I’m not looking to remove human governance entirely — just trying to understand whether the publish step is intentionally human-gated today.

[Ciara TN]

Hi @Okan Erdogan 

Great questions and it got me thinking of agents we have built. 

Giving Rovo skills allows it to do them with human intervention and won't do anything without the ok. 

If this is placed in an automation then the skill is removed and we use automation to publish. ie. Automation becomes the actor not the human. 

I think this is best practice as it doesn't let the agents run off and create 100s of pages without a human checking. 

-Ciara (Community Champion/Eficode)