Agent examples

Ask your agent across multiple apps

TWG CLI lets your agent pull context from multiple Atlassian apps in a single prompt, so you can get a complete picture without querying each app separately.

Multi-step workflow examples

These examples show how your agent chains multiple queries from a single prompt.

Example 1: Weekly work summary

What you say:

"Summarise everything I worked on this week."

What your agent does:

1. Looks up your recent Jira activity — issues updated, commented on, or transitioned.
2. Finds Confluence pages you've created or edited.
3. Checks pull requests you've opened or reviewed in Bitbucket.
4. Produces a plain-English summary you can share in a standup or check-in.

Example 2: Onboard to a new project

What you say:

"I just joined the Platform Engineering team. Help me get up to speed."

What your agent does:

1. Finds the team structure and key people.
2. Looks up the current active sprint and open issues.
3. Finds relevant Confluence docs and onboarding pages.
4. Lists recent pull requests in the team's repos.
5. Summarises what the team is currently focused on.

Example 3: Explore a Jira issue in depth

What you say:

"Tell me everything about PROJ-123."

What your agent does:

1. Fetches the issue details, current status, and assignee.
2. Looks up linked issues and dependencies.
3. Finds related pull requests and any linked Confluence pages.
4. Gives you a full picture of the issue and what needs to happen next.

Example 4: Prepare for a sprint review

What you say:

"Help me prepare for our sprint review. The sprint ends Friday."

What your agent does:

1. Lists all issues in the current sprint, grouped by status.
2. Flags anything overdue or not started.
3. Summarises completed work in a format you can share.
4. Identifies any blockers or at-risk items.

People across connected directories

Check the commands and selectors in your installed version. Reporting requires the enriched surface. Mixed-identity hierarchy and 3P activity/context also require the third-party identity feature for your caller and tenant context. Signing in does not itself enable it. Connected-provider data and permissions determine coverage; do not replace an unavailable route with another person's data.

Example 5: Prepare for a manager conversation

What you say:

"Find my immediate manager, then summarise their work and documents from the last two weeks, including supported connected-directory activity."

What your agent does:

1. Resolves your AAID and uses user manager --identifier on the selected site.
2. Uses data.managerIdentifier.ari for a 3P manager, otherwise data.manager.accountId for a 1P manager. Stops if neither is returned.
3. Passes that returned identifier to work/docs queries with --identifier, preserving the site and 14-day window. For bounded manager levels, uses user manager-chain --identifier --depth 2 when advertised and enabled.
4. Summarises returned evidence with links and any coverage limitations.

Example 6: Summarise direct reports' work

What you say:

"What did my direct reports work on over the last two weeks? Tell me if directory-only people are outside the available coverage."

What your agent does:

1. Resolves your account ID, then calls user direct-reports --identifier when the mixed-identity route is advertised and enabled.
2. Preserves each report's complete actorIdentifiers in one downstream query, never merging different people's actors. Reports an unavailable gate or a roster-bound failure rather than claiming a complete first-party-only result.
3. Keeps each person's identity separate in downstream work queries. If a full roster would require excessive calls, agrees a smaller scope or clearly states which people were sampled.
4. Groups evidence by person and outcome, preserving failures and missing coverage rather than silently dropping people.

Example 7: Find a person's meetings and videos

What you say:

"For this resolved person, find meetings and videos relevant to their work this fortnight."

What your agent does:

1. Preserves the person's verified AAID or returned identity ARI.
2. Checks the installed meetings/video query contract and 3P availability if needed.
3. Queries meetings/videos using --identifier, the selected site, and a 14-day window; it does not infer a reporting-peer roster from collaborators.
4. Uses supported provider-specific detail reads for selected results. It does not infer that every returned video has a retrievable transcript or that meeting attendance implies ownership.

Example 8: Explore a directory-only person's context

What you say:

"For this directory user's returned identity, show their conversations and mentions. Can you also rank their collaborators?"

What your agent does:

1. Checks context user help for third-party support, then preserves the returned identity in context user <3p-ari> with the selected site and time window.
2. Reports the supported conversation-membership and message-mention relationships, keeping target links and pagination limits visible.
3. Explains that collaborators requires a verified Atlassian account ID and cannot rank collaborators for a 3P-only subject. It does not substitute your account or label reporting peers or conversation members as ranked collaborators.

Example 9: Separate reporting peers from collaborators

"Who reports to the same manager as me, and who are my recommended collaborators? Keep those groups separate."

The agent uses user peers --identifier for the bounded reporting group and collaborators --account-id for the verified 1P subject's recommendations. It keeps the groups and scores separate and reports each route's coverage.

Example 10: Continue a partial document result

"Find my manager's documents from the last two weeks. If only a page is returned, continue once and explain what remains incomplete."

The agent preserves the manager's identity or complete actor set, site, and window across docs query calls. It follows only a returned cursor when hasNextPage is true, retains usable partial data, and reports remaining limits. An authorization denial stops that route, not an automatic login retry.

Example prompts quick reference

Use these prompts as starting points. Your agent understands natural language, so you don't need the exact wording.

Next steps

• Install skills