When an external agent discovers Rovo via the Agent Card, it sees a list of skills - high-level capabilities that describe what Rovo can do. Unlike MCP tools, which are individual API operations, A2A skills represent broader task categories that Rovo can handle through natural language.

How skills work

Skills are advertised in the Agent Card's skills array. Each skill has:

id — A unique identifier for the skill.
name — A human-readable name.
description — What the skill does and when to use it.
tags — Keywords for categorization and discovery.
examples — Example prompts that demonstrate the skill.

External agents use this information to decide when to delegate tasks to Rovo. See Supported JSON-RPC methods for the methods the gateway accepts. Use message/stream for any request that may run longer than approximately 30 seconds.

The skills below reflect the static fallback Agent Card built into the gateway. When the downstream Rovo service is available, the live Agent Card may advertise different or additional skills. Always inspect the live Agent Card at https://a2a.atlassian.com/.well-known/agent.json for current values.

Available skills (static fallback card)

Work in Jira

Field	Value
ID	`work-in-jira`
Description	Get Jira context like project status, work item details, or sprint data right where you're doing analysis. You can also create, edit, delete, and update work items in your Jira spaces.
Tags	`Jira work item`, `sprint backlog`, `sprint planning`, `kanban`, `scrum`, `project planning`, `project management`

Example prompts:

"Summarize the status of all open Jira work items assigned to me. Include ticket links and context ordered by priority and deadline, highlighting any critical details."
"List my Jira issues in "Blocked" status, ordered by due date. Include the reason they are blocked, the last update made, and their project links."
"Create a new Jira work item in the project [project name] based on this project brief [project brief link]. The work item should include a summary, description, acceptance criteria, and priority."
"Provide a list of tips to maximize productivity when using Jira across different Atlassian apps. Include detailed examples and prompts, ensuring each one specifies a clear goal, relevant context, and expected outcome."

Work in Confluence

Field	Value
ID	`work-in-confluence`
Description	Find answers from Confluence and create pages. Turn meeting notes, specs, or drafts into team documentation instantly. You can also edit, delete, and update existing Confluence pages.
Tags	`Confluence space`, `knowledge base`, `release notes`, `meeting notes`, `product specs`, `documentation`

Example prompts:

"Draft a report for Confluence summarizing the key updates and additions I made over the past week. For each significant update or addition, create a subsection with a brief explanation, focusing only on the essential information"
"Create a Confluence page with all open Jira tickets assigned to me, ordered by priority and deadline. For each ticket, include a direct link, priority level (High, Medium, Low), deadline, and any relevant context or updates."
"Find Confluence pages related to [customer name]. Summarize each of the pages, highlighting the main ideas and key points that would be important for my upcoming conversation."

Dynamic vs. static skills

The Agent Card, and therefore its skills list, can come from two sources:

Dynamic — Fetched from the downstream Rovo service, which provides the canonical agent identity and skills. The gateway caches this for 5 minutes and hydrates it with local auth config.
Static fallback — If the downstream service is unavailable, the gateway serves a hardcoded Agent Card with the skills listed above.

Skills vs. MCP tools

Aspect	A2A Skills	MCP Tools
Granularity	High-level task descriptions	Individual API operations
Invocation	Natural language messages	Structured tool calls with parameters
Discovery	Agent Card `skills` array	MCP `tools/list` method
Example	"Work in Jira"	`searchJiraIssuesUsingJql(jql, cloudId)`

Input and output modes

The static fallback card advertises:

Input modes: text/plain
Output modes: text/plain

The live dynamic card may advertise different modes. The Rovo Agent Adapter currently extracts text from text parts only — file and data parts are accepted but not processed. Support for additional content types may be added in future releases.

Next steps

Architecture — Review the Agent Card fields and caching behavior.
Getting started — Send your first message to Rovo.
Authentication — Learn how agents authenticate before invoking these skills.