Try this skill in Strawberry

Set Up a Shared Team Workflow

Turn a workflow that already works into a method the team can trust and repeat. Start from a real accepted example. Do not use this skill to write generic process documentation or redesign a broad operation before anyone has proved the underlying work.

1. Start from the working method

Identify the workflow, the result it produces, the people who use or depend on it, and the example that proved useful. Reconstruct how the result was actually made across messages, files, tabs, documents, project tools, connected apps, and human judgment.

If there is no accepted example yet, help the user complete or calibrate one focused workflow first. A shared method should preserve what worked, not turn an untested idea into team policy.

2. Decide what the team needs to share

Separate the durable method from one person's private context and preferences. Agree on:

  • the trigger and intended result;
  • required inputs and their source of truth;
  • the owner, contributors, reviewers, and handoff;
  • the shared destination and useful source links;
  • choices the companion may make versus choices a person must review;
  • privacy, client, candidate, employee, or other context boundaries; and
  • the exceptions that should stop or reroute the work.

Choose the smallest sharing layer that solves the problem. Share the approved artifact when the team only needs this result. Create a team skill when they need the method. Share the full companion only when teammates need its ongoing approved context, memory, and several workflows. These are separate choices.

3. Build the reusable team method

Turn the accepted workflow into concise team guidance another companion or teammate can follow without guessing. Preserve the judgment, source priority, output, review points, and escalation logic that made the original work trustworthy. Do not copy personal notes, incidental steps, or temporary workarounds into the shared version.

Use the team's real workspace. Strawberry can follow the workflow across visible tabs, logged-in sites, shared files, communication tools, and approved apps while the user inspects or takes over. Put the team skill and shared artifacts where the intended people can use them, and verify that its links, permissions, fields, and destinations work for someone other than the original owner.

4. Pilot one real handoff

Run the method on a representative piece of work with the people who will own and review it. Check whether they can find the inputs, understand the result, correct the companion's judgment, approve the right actions, and pick up the next step.

Keep the pilot small enough to correct. Surface unclear ownership, unavailable context, permission gaps, conflicting source meanings, and steps that depend on tribal knowledge. Update the team skill from the accepted corrections, then rerun the handoff that failed before expanding.

5. Define the Routine precisely

Once the pilot works, define the concrete Routine that carries the method forward:

  • Trigger: the event, state change, or cadence that starts it;
  • Systems checked: the exact approved sources it reads;
  • Actions: what it gathers, compares, drafts, creates, or proposes;
  • Output: where the result appears and who reviews it;
  • Approval behavior: which messages, record changes, permissions, or other writes wait for a person; and
  • Stop conditions: missing identity, conflicting sources, unavailable permissions, sensitive context, or another exception that needs judgment.

Do not enable recurrence merely because a schedule is available. Keep the Routine as a reviewed proposal when the trigger is weak, the method is still changing, or on-request use is enough.

6. Hand over and improve the workflow

Deliver the verified team skill, shared destination, ownership map, pilot result, and Routine plan together. Make remaining gaps and temporary manual steps visible.

Use feedback from real runs to improve the method without allowing one exception to rewrite it. Review ownership, permissions, source meanings, and approval behavior when the team or systems change. Keep sharing the method, inviting teammates, enabling the Routine, and making external changes as distinct approved actions.

A useful AI workflow can easily stay personal. One person knows which sources to trust, how to judge an exception, and what still needs review. A teammate sees the output and has to reconstruct the method.

A generic process document rarely fixes that. What the team needs is the judgment and the handoffs that made the result reliable, without copying one person's private notes.

Strawberry can keep that method as a team skill, connect it to approved shared context, and run it across the tools the team already uses. Start with one workflow and check that someone else can complete a real handoff.

Start with a workflow that already works

Choose a result that one person has already produced and corrected. The accepted example reveals the real inputs, source priority, judgment, output, review points, and exceptions. Without that evidence, a shared workflow is only a proposed policy.

If the underlying result has not been tested yet, complete or calibrate the focused workflow first. The team should preserve what proved useful rather than standardize an assumption.

Share the smallest useful layer

Sometimes the team only needs the approved artifact. Sometimes several people need the method, so a team skill is the right layer. A shared companion is useful when the team needs ongoing approved context, memory, and several connected workflows. Those are different choices, and a working personal method does not need to become a company-wide system.

The reusable guidance should contain the trigger, inputs, source of truth, owners, reviewers, destination, decisions the companion may make, actions a person must approve, and exceptions that stop the work. Personal preferences, client-confidential context, and temporary workarounds stay outside unless they are deliberately approved for the team.

Build and test it in the real workspace

Strawberry can follow the workflow through visible tabs, logged-in sites, shared files, communication tools, and connected apps while the user watches or takes over. That makes it possible to test more than the written instructions: whether links and permissions work, field meanings are shared, the destination is correct, and the next owner can actually pick up the result.

Pilot one representative handoff with the people who will use and review it. A small run exposes missing access, tribal knowledge, unclear ownership, and review steps that looked obvious to the original operator. Correct the team skill and rerun the handoff that failed before broadening it.

Add a Routine when the method is ready

A useful Routine is more specific than a schedule. It names the event or cadence that starts the work, the approved systems it checks, what it gathers or prepares, where the result appears, who reviews it, which external changes wait for approval, and the conditions that stop the run.

Recurrence should come after the team trusts the handoff. When the trigger is weak, the method is still changing, or on-request use is enough, keep the Routine as a proposal. Team sharing, enabling the Routine, inviting people, and changing external systems remain separate decisions.

The finished handoff includes the verified team skill, shared destination, ownership map, pilot result, and Routine plan. Later runs can improve the method as tools and teams change without allowing one exception to rewrite the whole workflow.