Rovo Agent

The rovo:agent module defines an Agent. Agents are configurable AI teammates that integrate into Jira and Confluence workflows. You can define an Agent's behavior using a prompt and an action, so the Agent can fetch data and perform operations.

Data access

Unlike Agents built using an Atlassian app user interface (UI), app-based Agents only have access to the data in the workspace that the app is installed in. For instance, if there is a Confluence page located at yourtenant.atlassian.net that describes your team's CI/CD process, an app-based Agent that is installed in Jira at yourtenant.atlassian.net will not have automatic access to it.

To enable your Agent to access data from multiple Atlassian apps, configure your app to support multiple Atlassian apps. See App compatibility.

Manifest structure

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
modules {}
└─ rovo:agent []
   ├─ key (string) [Mandatory]
   ├─ name (string) [Mandatory]
   ├─ description (string) [Optional]
   ├─ icon (string) [Optional]
   ├─ prompt (string) [Mandatory]
   ├─ conversationStarters [] [Optional]
   │  └─ conversationStarter (string)
   ├─ actions [] [Optional]
   │  └─ action (string)
   ├─ skills [] [Optional]
   │  └─ skill (string)
   └─ followUpPrompt (string) [Optional]

resources []
└─ key (string) [Mandatory]
└─ path (string) [Mandatory]

In this structure:

• The rovo:agent array includes properties such as key, name, description, icon, prompt, conversationStarters, actions, skills, and followUpPrompt.
• The conversationStarters, actions, and skills arrays are represented with generic entries conversationStarter, action, and skill, respectively.
• The resources array includes properties key and path.

Properties

Manifest example

Here is an example manifest file for creating an Agent that assists with managing project risks:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
modules:
  rovo:agent: 
    - key: risk-agent
      name: "My Risk Register Assistant"
      description: A Rovo Agent that helps you manage your project risks
      icon: resource:example-resource;icons/risk-agent.svg
      prompt: |
        You are a helpful assistant that helps users manage their project risks. 
        You can retrieve risks from the risk register, create new risks and update existing ones.
      conversationStarters:
        - Fetch my active project risks
        - Fetch my highest priority risks
        - Create a new risk
      actions:
        - fetch-all-risks
        - fetch-risk-by-priority
        - create-risk
        - update-risk
resources:
  - key: example-resource
    path: static/hello-world/build

prompt as a resource

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
modules:
  rovo:agent: 
    - key: risk-agent
      name: "My Risk Register Assistant"
      description: A Rovo Agent that helps you manage your project risks
      icon: resource:example-resource;icons/risk-agent.svg
      prompt: resource:agent-resource;prompts/agent-prompt.txt
      conversationStarters:
        - Fetch my active project risks
        - Fetch my highest priority risks
        - Create a new risk
      actions:
        - fetch-all-risks
        - fetch-risk-by-priority
        - create-risk
        - update-risk
resources:
  - key: example-resource
    path: static/hello-world/build
  - key: agent-resource
    path: resource/agent

Here agent:prompt refers to following string resource: resource/agent/prompts/agent-prompt.txt

1
2
3
You are a helpful assistant that helps users manage their project risks. 
You can retrieve risks from the risk register, create new risks and update existing ones.

Agent interaction points

Programmatic access via Forge bridge

Using the Forge bridge rovo API, you can initiate a conversation in the Rovo chat sidebar with your Forge Rovo agent.

Chat side panel (Confluence and Jira)

Accessed by clicking the Chat button in the top navigation bar

AI toolbar in the Confluence editor

Accessed using the /ai command in the editor

The AI toolbar in the Jira editor

Accessed using the /ai command in the Jira issues editor

Automation (Confluence and Jira)

You can add Agents to Automation rules. This will invoke the Agent to act asynchronously in response to Atlassian app events or schedules. When users configure an automation rule they will set an additional prompt with specific instructions how to act during that rule. The response from the Agent can be passed to subsequent steps in the automation rule using smart values.

Choose between prompts, skills, and actions

Keep your Agent's prompt focused on its overall role, behavior, boundaries, and high-level responsibilities.

Use a rovo:skill when instructions apply to a particular type of task rather than every conversation. Skills are especially useful when a workflow:

• Coordinates multiple actions or calls them in a particular order.
• Contains branching steps, validation, or error-recovery guidance.
• Requires detailed examples or supporting reference material.
• Is reusable across multiple Agents.
• Would otherwise make the Agent prompt difficult to understand or maintain.

Skills provide task-specific instructions that an Agent can load when relevant. Actions provide the underlying executable capabilities, while the skill explains how to combine those actions into a complete workflow.

Don't use skills only to divide Agent-wide instructions into smaller files. If instructions must apply to every interaction, keep them in the Agent prompt or store the prompt as a resource.

Writing effective prompts

Define the Agent's purpose, personality, output format, capabilities, and other relevant aspects. Include these key components:

Define the role of your Agent

The role you define shapes the Agent's language, tone, style, and personality.

Example:

1
2
You are an expert project manager tasked with managing risks for a project.

Outline what your Agent can assist with

Outline the various jobs that the Agent can assist users with. The list does not need to be comprehensive.

1
2
3
4
5
6
7
You can help with the following jobs:

A. Reviewing high priority risks
B. Generating executive reports on project risks
C. Updating the risk register with new risks
D. Evaluating the impact of risks on project objectives

Format and structure your prompt

Add logic/structure to longer prompts using delimiters. This improves instruction quality and readability.

1
2
3
4
5
6
7
8
9
10
11
I'll separate the instructions for each job with a. '---' on a new line, followed by the job title.
---
A. Reviewing high priority risks
When asked, you can help teams to review high priority risk and critique it based on likelihood, impact, and mitigation.
To do this, follow these steps:

--- 
B. Generating executive reports on project risks
To do this, follow these steps:

Define when actions should be invoked

Instruct your Agent on how to execute more complex jobs by defining the actions it should take.

1
2
3
4
5
6
7
8
Reviewing high priority risks
When asked, you can help teams to review high priority risk and critique it based on likelihood, impact, and mitigation.

To do this, follow these steps:
1. Retrieve the list of high priority risks from the risk register using the fetch-risk-by-priority action.
2. Critique each risk based on likelihood, impact, and mitigation.
3. Provide a final rating of the risk on a scale from 1 to 10, where 1 is an extremely low risk and 10 is an extremely high risk.

Define the format of output

Define how your Agent will structure its responses when performing certain jobs.

1
2
3
4
5
* Use emoji to highlight what the rating score for each risk. Use this emoji scoring
🔴 - Red circle emoji to highlight if there is a high risk (8-10). 
🟡 - Yellow circle emoji to highlight if there is a medium risk (4-7)
🟢 - Green circle emoji - to reflect low risks (1-3). 

These are some basic guidelines to help you craft effective prompts for your Agent. You can customize these guidelines to suit your Agent's specific requirements.

Tutorial