Hey all!
I had a great public class on best practices for prompting earlier today (there's more under Learning → Public Classes!).
This is a fun one for me since we focus SOLELY on theory - what makes a good prompt? What doesn't? When should we split up a prompt?
I'm particularly interested in how folks address question #2 (below) - how do YOU save your preferences for Rovo?…
Check out the questions below, and review Pubic Classes for future sessions!
Question 1 - Can Rovo use source documentation located beyond the Atlassian environment (e.g., located in a Box folder)?
- Yes, Rovo can access external documents if your organization’s administrator connects third-party tools (such as Box, Google Drive, or SharePoint) into the Atlassian Teamwork Graph via supported connectors. Rovo respects existing user permissions and will only display content you are authorized to see. Alternatively, users can paste document links directly or manually upload files into a session prompt if supported and permitted by corporate security policies. Note that connectors tend to be the preferred method1
- Source -
https://www.atlassian.com/software/rovo/connectors
Question 2 - Can I point Rovo to a complex specification file (such as YAML or Markdown) or an entire folder containing multiple documents to drive multi-phase tasks?
- While local file folder trees cannot be mounted directly as local directories inside the cloud UI, you can store your specs across Confluence pages or connected enterprise cloud storage (e.g., Box/Drive) and reference those URLs directly inside your prompts. You can also upload markdown/text files into supported chat attachments per prompt, or use Rovo’s Memory and Custom Agent definition settings to embed system instructions and reference frameworks persistently.
Question 3 - Are there standardized tags for Jira-specific items in Rovo (for example, finding tickets less than 5 weeks old without relying strictly on ambiguous plain text)?
- Yes. Rather than relying solely on natural language interpretations, you can prompt Rovo using Jira Query Language (JQL) syntax or explicitly ask Rovo to generate the exact JQL string (e.g.,
created >= -5w). Rovo parses JQL parameters directly, displays the generated search query syntax, and returns a verified URL link pointing straight to the filtered Jira issue list.
