FROM A REQUEST TO AN ARTIFACT
Give your agent
a brief.
A workspace connects a model to files, tools, and feedback so work can continue across several steps.
Make the purpose, plan, and result visible
An agent workflow can propose an action, use a tool, inspect the result, and choose what to do next. For faculty, that might mean creating a lesson page, opening it, identifying a broken interaction, and revising it. The teaching brief gives this loop a purpose and a stopping point.
Start with a brief you can use.
Fill in a few decisions and assemble a Markdown brief for Cursor, Antigravity, or another workspace. This builder uses a template; your entries stay in this page.
Your course brief
Your brief will appear here. Complete the fields or load the campus example, then review the result before using it.
Save the brief alongside approved readings and data. Ask the agent for a plan before it builds.
Begin with a concrete brief
Describe the learners, intended performance, approved inputs, desired resource, and review criteria. Keep the task small enough to inspect. Our campus brief asks first-year students to compare proposals using supplied evidence; it also explains that the scenario and data are invented. Those boundaries help a draft stay connected to the teaching purpose. Use the brief builder to turn a need into a request you can save. Then ask the agent to summarize the requirements it will follow. This gives you an early opportunity to correct a misunderstanding before it becomes several finished files.
Review the plan and open the result
Cursor and Antigravity are examples of environments that support work on reviewable artifacts. Cursor documents a plan-and-build workflow. Antigravity offers artifacts and configurable review policies; its current standalone application differs from Antigravity IDE. Exact controls and capabilities depend on the installation and account. Watch for the transferable process: identify the folder, supply the brief, examine the plan, and inspect what was created. If the tool cannot produce a requested presentation format, a readable document or browser deck may be a useful alternative. A completion message is the beginning of your inspection, not evidence that every learner action works.
Divide responsibility, then connect the findings
Assign a researcher to map claims to approved material, a writer to draft instructions, and a critic to compare the draft with the outcome. Give each a distinct output and the same brief. Supported subagents can separate work; separate conversations or sequential passes can also provide useful review stages. Role names do not create expertise, and repeated agreement can preserve a shared error. Check a criticism against the source or artifact before accepting it. Keep drafts, findings, and accepted versions together so the next revision can begin with the reasoning behind the resource.
PUT THE IDEA TO WORK
Try this in your course.
Cursor: plan first
Ask the agent to plan one small resource from your brief. Review the plan before asking it to build.
Antigravity: review the work
Give the agent a bounded task. Inspect its plan, files, and preview before accepting the result.
Give the agent a brief
Write the course context, approved sources, constraints, and review criteria in one reusable file.
Give each role a job
Assign a researcher, writer, and critic different tasks. Give each role an output and a review criterion.