How to Delegate Work From Claude Code to Jira and Slack

To delegate work from Claude Code, turn the session into a task someone else can execute: the outcome, relevant evidence, steps, acceptance checks and a clear owner. The delegate-to-teammate skill in Rhize Ops can prepare that package and, when authorized, deliver it through your configured Jira and Slack connections.

What does a useful AI-to-teammate handoff contain?

A useful handoff answers what needs to happen, why it matters, which constraints apply and how the recipient will verify completion. A transcript alone is usually too much context; a one-line assignment is often too little. The skill is designed to connect a concise tracker issue to the supporting material.

The workflow grew from our own delegation work, but recipient identities and project mappings belong in private configuration. It does not require a colleague with a particular name or Rhize's internal channel structure.

Where does delegation lose context?

The common gap is between a decision made in an assistant session and the task record another person receives. The package should preserve the decision and its evidence without exporting unrelated chat, private filesystem paths or unnecessary customer information.

Write a measurable done condition before creating the issue. For a sitemap repair, for example: identify the broken URL, explain the expected canonical URL, make the scoped fix and verify the public result. Include the known constraint and the relevant source; do not make the teammate reconstruct the whole session.

When several tasks are involved, confirm each project and owner separately. A shared conversation does not mean all tasks belong in the same Jira project or have the same due date.

What does the current operations plugin include?

The current Rhize Ops release includes delegation, decision pre-mortems, cross-model handoffs, parallel-agent coordination and a skill-usage dashboard. This guide focuses on delegation. The other capabilities are separate workflows, not prerequisites for assigning a task.

The skill-usage dashboard reads a separate rhize-skill-monitor checkout. It is useful for reviewing installed-skill activity, but is not needed to create a Jira/Slack handoff.

Set up the recipient and destinations once

Install rhize-ops from the marketplace, then run /rhize-ops:delegate-setup. The setup resolves recipient, tracker and Slack identifiers through connected tools and writes a private JSON configuration outside the plugin. Confirm the recipient and channel instead of copying another team's identifiers.

The current configuration lives at ~/.claude/rhize-ops/delegate.config.json. It holds the recipient, Jira IDs, Slack channel and project mapping, with an optional Confluence destination. Missing integrations are marked incomplete. Rerun setup when an owner, project or destination changes; do not put credentials in a public repository.

Before delivery, review which supporting files are appropriate to share. The workflow can export scrubbed vault notes and attachments when the required access is configured. Unavailable files belong in a request list, not as broken local paths in a message.

What gets written to Jira, Confluence and Slack?

The Jira issue gets a concise description with the outcome, done criteria, a key gotcha and a starter prompt. If configured, a Confluence page holds the fuller task package. Supporting material can be attached to Jira and shared through Slack Canvases. The recipient receives a tagged Slack message with task details in its thread.

The current protocol gives each approved task a stable ID before external writes. That ID connects the Jira description and Slack task reply. If Jira creation is unavailable or unconfirmed, the handoff must say needs_jira instead of inventing an issue or silently retrying an ambiguous creation.

Verify the actual issue and message after writing. A successful tool call is not enough if the wrong project, recipient or attachment was selected. If an integration fails, preserve the confirmed parts and report what remains incomplete so the next attempt can resume without duplicates.

A prompt you can adapt

Supply the target outcome, recipient, project, due date and authorized source material. If any of those are unknown, leave them open for clarification rather than making them up.

• "Prepare a handoff for Alex in our configured website project: repair the sitemap entry for the retired service page. Include the observed URL, the expected redirect, the relevant notes and a test of the public result. Use Friday as the proposed due date. Show me the package before creating the issue and notifying Alex." This is an illustrative prompt, not an instruction to contact a real person.
• Once the package is approved, ask the assistant to deliver the same reviewed task through the configured Jira and Slack destinations. Review the links and readback so the recipient can reach the evidence.
• For a separate skill-usage review, ask for the skill dashboard. Its empty state can mean no snapshots are available; it does not establish that nobody used a skill. Consult the separate monitor setup instructions rather than an obsolete bundled script path.

When is this workflow useful?

It fits teams that hand work from an assistant session to another person and already use a tracker and team messaging. If a task can be completed in one short exchange, a smaller handoff may be enough. Match the package to the complexity of the work.

The intended benefit is less reconstruction by the recipient and clearer completion criteria. Measure clarification requests, missing attachments and rework before claiming a time saving. A more detailed task is not automatically a better task.

Start with one bounded handoff

In Claude Code, use /plugin marketplace add https://github.com/Rhize-Media/rhize-plugins, then /plugin install rhize-ops@rhize-plugins and start a fresh session. Run /rhize-ops:delegate-setup, verify the configured destinations, and ask for a proposed handoff before authorizing external writes.

The current setup, attachment requirements and fallback behavior are documented at https://github.com/Rhize-Media/rhize-plugins/tree/main/rhize-ops. Start with one task whose owner and done condition are clear. Use the Rhize Plugins marketplace to try the workflow with your own configuration.