// Guide

Common workflows

Choose an example that matches your work. Each workflow combines the same building blocks: source material, a clear task, a place for discussion, and a decision about the result.

What is AI artifact management?

AI artifact management means organizing AI-generated work, recording its sources and versions, and deciding which result a team can use. In this guide, an AI artifact is a draft, plan, policy, or other document that an agent creates or changes.

For example, an agent proposes a change to a governed launch checklist in ArtifactBridge. A person reviews the diff and accepts it or requests changes. The accepted managed-document version keeps attribution and history. Another agent must read that version before reporting that it has the approved rule.

Choose where to keep the work

Document approval does not update CLAUDE.md or AGENTS.md in a repository. Keep those file changes in the repository review process. See how approval works for document review settings.

Verify the result before a handoff

1. Name the source document and the version that the agent read.
2. Keep the proposed edit distinct from the accepted document version.
3. After acceptance, ask a separate agent to read the accepted version and check the requested rule.
4. Record what the check proves and what remains untested.

The completed public practice review includes an accepted sample policy, a separate agent readback, and a JSON record with the content hash. It verifies one sample document change. It does not prove that later tasks follow the rule.

Copy this checklist into the task handoff and fill in each field.

Build a project knowledge base

1. Create a project folder in the Library.
2. Import the current plan, decisions, specifications, and other relevant material.
3. Check titles, sources, access, and review settings.
4. Copy folder context into your team’s connected AI projects.
5. Ask each agent to read the live folder at the start of a task.

Result: different tools can refer to the same workspace material. Check source freshness and document versions when decisions depend on recent changes.

Draft and review a team document

1. Create a governed document and attach the approved source material.
2. Ask an agent to propose a draft with citations and explicit assumptions.
3. Review the change in the Inbox.
4. Request targeted changes or accept the result.
5. Share the approved document with its intended readers.

Result: the document has a reviewed version and a record of the decision. Use a working document instead when you explicitly want direct agent updates during drafting.

Take an investigation to a local coding agent

1. Open a Room for the issue and attach the relevant specification or evidence.
2. Use Send to agent to start or resume your local coding tool.
3. Give the tool the repository and the permitted scope of code changes.
4. Ask it to record findings and its final result in the Room.
5. Review code through your repository workflow and document changes through ArtifactBridge.

Result: the team retains the reasoning and outcome while the local tool works with the repository. Room approval is not a substitute for a code review or merge decision.

Bring a local project into the Library

1. Connect a local agent through ArtifactBridge Desktop.
2. Open Upload → Ask your agent and enter the project folder.
3. Ask for the useful project documents and explain what to exclude.
4. Inspect the import proposal’s scope, titles, destinations, and content.
5. Approve the intended import and check the resulting Library folder.

Result: the selected knowledge becomes available to the team without treating the entire local directory as approved content.

Run a weekly report

1. Write an instruction document that names the sources and report structure.
2. Create a workflow using that document.
3. Select the weekly day, hour, and time zone.
4. Choose a target folder and optional Slack channel.
5. Activate it, inspect the next run, and check the first result in run history.

Result: a repeatable report with a visible owner and run record. Review failures and incomplete sources before relying on the report.

Onboard a teammate and their AI tool

1. Invite the teammate and confirm the required role.
2. Grant access to the project’s private documents or Rooms if needed.
3. Have them connect their own AI tool and verify the workspace.
4. Share the project folder context and relevant Skills.
5. Complete one small proposed change together.

Result: the teammate and their agent can find the right material and follow the team’s review process.

Review a customer’s Word changes

1. Identify the document version sent to the customer.
2. Add the returned Word redline against that document or version.
3. Decide each edit and comment on rejections.
4. Accept the completed review, then download the reply file.
5. Inspect and send the file with its decision summary.

Result: the decision record stays with the document, and you have a reviewed file to return.

Share approved knowledge outside the workspace

1. Confirm that the document and referenced images are ready for external readers.
2. Review its sharing settings and enable public access only if intended.
3. Check the link’s version behavior.
4. Verify the public reading view, then send the link.
5. Disable public access or revoke the relevant link when it is no longer needed.

Result: external readers can see the intended content without gaining document editing rights.