ROVO Agent - It always reads from an OLD confluence source.

[Krishna Modem]

Bleo

"ACTIVATE SOURCE-ONLY MODE:

1. Zero-Cache: You are forbidden from using any data (OS names, hostnames, dates) from previous turns or your internal memory.

2. Live-Scan Only: For every single response, you must execute a fresh scan of the provided URL.

3. Strict Verification: If a piece of data is not present in the current tool output for the provided URL, you must report it as 'Not Found'.

4. No Assumptions: Do not 'fill in the blanks' or guess OS versions. If the source says 'Ubuntu', do not write 'Ubuntu 22.04' unless the version is explicitly in the table cell.

5. Audit Trail: Before presenting a table, internally verify that every row matches the latest scan result."

---

Phase 1: Create User Stories

1. Targeted Page Scan: Access the Confluence page provided. Locate the database or table and extract ONLY the following fields: OS, name, and overdue_date.

2. Global Scan: Scan the entire table to identify every unique OS group present.

3. Cross-Reference Jira: For every identified unique OS , check if a Story exists in STRATOSBAU (Summary: [OS]).

4. Pre-Flight Validation Report: Present the data in the following section:

5. Conditional Call to Action:

   • Scenario A: If NO NEW User Stories are needed (All are DUPLICATES):
     Present the message: "All identified user stories are present."
     Followed by:
     "To proceed with creating the Sub Task, please choose:

     • Type 'Proceed': To Progress with Phase 2 (Sub-Task Validation)

     • Type 'Cancel': I close the session and clear data."

   • Scenario B: If NEW User Stories need to be created:
     Present the message:
     "To proceed with creating the User Stories, please choose:

     • Type 'Proceed': To create User Stories

     • Type 'Cancel': I will clear this data"

6. Execution for Scenario B (Only after 'Proceed'):

   • Create Story:

     • Summary: [OS]

     • Description: Vulnerability Remediation for hosts.

     • Labels: THREAT

   • Final Phase 1 Table: Once all groups are finished, provide a final table:

     • OS Group | Story Key |

   • Next Step: After displaying the table, ask the user to Type 'Proceed' to begin Phase 2 (Sub-Task Validation).

[Rebekka Heilmann _viadee_]

Hi @Krishna Modem 

what exactly is your problem? What do you mean by "Old" Confluence source?

[Krishna Modem]

I tested this agent initially using a Confluence page as the data source, and everything worked as expected.

However, when I updated the agent to use a different Confluence page containing a new set of data and re-ran it, the agent continued to return results from the original Confluence page.

To troubleshoot, I tried the following:

• Updated the source to the new Confluence page
• Created a new version of the agent
• Removed/deleted the original Confluence page
• Re-ran the agent multiple times

Despite these changes, the agent still appears to be referencing the content from the original page and returns answers based on that data.

[Arkadiusz Wroblewski]

Hello @Krishna Modem 

Did you just Deleted Page or also Purged it from trash to delete Completely ? If you sended it only to trash it can theoritically contextual stay avaiable for Rovo use.

Use Debug Response Debug responses from Rovo agents | Rovo | Atlassian Support

It should give you some answers

[Krishna Modem]

If i have give a new source and URL of confluence- why its not taking that as input ? How to resolve this?

[Arkadiusz Wroblewski]

@Krishna Modem 

And when you change configuration https://support.atlassian.com/rovo/docs/knowledge-sources-for-agents/ 

Have you tried changing the settings, for example, turning off and on the organization's knowledge, and adding your new pages as sources of knowledge?

Also, did you have permission to view it?