Clanker CloudA useful AI assistant often begins with one person's instructions and examples. Sharing it well lets colleagues reuse the method without accidentally carrying over private client material or assuming that every copy has the same connections.
Clanker Cloud's Team library workflow lets you publish a reviewed version of an Agent definition with its attached Skills. Team members can add editable copies to their own Workspace. The phrase “AI employee” describes this software role; it does not give a fictional persona real qualifications or independent responsibility.
Share a job that has already worked
Suppose an operations coordinator has built an Agent that turns weekly project notes into a leadership brief. It uses a consistent structure, identifies missing facts, and keeps confirmed decisions separate from suggestions.
Before sharing, run it on a sample the team can inspect. Save the expected output and explain what passed review. “It writes well” is difficult for a colleague to evaluate; “it preserves the reporting window and flags unassigned actions” describes behavior they can test.
The custom assistant walkthrough explains how to establish a role and test it with complete and incomplete inputs.
Review the definition and its Skills
Remove client names, private records, and credentials from reusable instructions. Replace local assumptions with explicit setup requirements. If a Skill expects a particular file structure, describe that structure without embedding a real customer's data.
Give the shared version a description that states what it receives, what it produces, and what colleagues must check. Mention required capabilities where relevant. A specialist that prepares a report is a different setup from one that also sends it to a distribution list.
Create or join the appropriate Team in your organization. From a saved Agent, use Share with Team, select the Team, review the definition and attached Skills, and publish the version through the available controls.
Explain what a colleague receives
The current workflow shares the definition and attached Skills as a version that members can add. It does not silently clone the creator's whole working environment.
| Item | What to expect |
|---|---|
| Agent definition and attached Skills | Included in the published version for members to review and add |
| Member's installed Agent | An editable copy that can be adapted |
| Connections, reference files, private memory, credentials | Require separate setup; not automatically transferred with the definition |
| Later published versions | Reviewed and added explicitly; do not silently overwrite personal edits |
For the reporting Agent, a teammate should add their own authorized Project files and check the tools they intend to use. Publishing the definition does not grant access to every client Project in the organization.
Onboard the first colleague with a sample task
Ask the first recipient to install the copy and run a fictional example. Compare the output with the expected brief. If it fails, check missing context and capability setup before changing the general instructions.
This small exercise exposes hidden dependencies: a reference file that only the creator had, a tool connection that was never documented, or a date format that a Skill assumed. Fix the shared setup notes so the next colleague can repeat the process.
Treat updates as changes to a working method
When you improve the Agent, describe what changed and which example you used to check it. Colleagues may have customized their copies, so they need enough information to decide whether and how to adopt the new version.
Authorized contributors and Team editors can manage shared versions. Archiving removes an item from the library; copies that members already added remain in their Workspaces. Do not treat archiving as revoking every installed copy.
If the organization needs a tailored workspace or rollout, the enterprise campaign describes that separate scoping path. For everyday adoption, start with one tested specialist and one colleague who can verify that it does the job.
Sources
Give your assistant one useful job
Publish a reviewed version of a useful Agent to your Team library so colleagues can add and adapt their own copies.