Build the Brief Before the Work

Apply context-cleanup judgment to a real-world task

What you will make

Create, sort, and test a context brief before asking a chatbot to complete a concrete, low-risk task. This activity applies the foundational maintenance moves from Context Cleanup: identify a source of truth, exclude unsafe or stale material, preserve uncertainty, and make a reviewable change log.

By the end, you can build a context brief for a real-world use case, explain why each source belongs in or outside the brief, compare two outputs after changing one meaningful condition, and name what still needs human verification.

Lab materials

Use during the lab

• Documents
  • Build the Brief Before the Work Required · Markdown · A structured record for the task, context inventory, brief, comparison, and maintenance decision.
• • Lab Notebook Interactive · An optional place to record the work and download a local copy.
• • Learner guide Markdown

Go deeper

• Related articles
  • Build the Brief Before the Work Article
• • Gallery Gallery
• • Facilitator guide Markdown

Before you start

Level and format: Beginner; individual, pair, or small group

You need: a low-risk use case, worksheet, and chat tool if you want to run the output comparison. Use fictional, volunteered, or safely redacted material. Do not use confidential business, client, medical, legal, financial, employment, or other consequential decisions.

The core idea

Context Cleanup teaches the reusable maintenance skill. This lab asks what that judgment changes when a tool helps with a specific task: an invitation, FAQ, orientation, meeting agenda, or other bounded real-world artifact.

Build, test, and maintain the brief

Complete the activity here. Open the optional Lab Notebook beside this page when you want a place to record your work.

1. Choose a Low-Risk Task

   Choose a coffeehouse announcement, volunteer orientation, meeting agenda, short presentation, personal project outline, simple fictional FAQ, or community activity. Start from the demonstration examples if you need a safely bounded use case.

   Avoid tasks requiring sensitive personal, legal, medical, financial, employment, or confidential business information.

   Open the context brief worksheet and use it throughout the remaining steps.

2. Create a Context Inventory

   List:

   • goal
   • audience
   • source material
   • constraints
   • active decisions
   • preferences
   • examples
   • unknowns
   • verification needs
   • information that should not be shared

3. Sort the Context

   Place each item into one category:

   • Required - the task cannot be done appropriately without it.
   • Useful - materially improves the result.
   • Optional - may help but is not needed now.
   • Exclude - irrelevant, sensitive, stale, conflicting, or distracting.
   • Verify - potentially important but not yet trustworthy or current.

4. Identify the Source of Truth

   Name the authoritative source, supporting references, informal notes, and assumptions. If there is no authoritative source, record that uncertainty.

5. Give the Chatbot the Brief

   Restate the goal, audience, active constraints, source of truth, assumptions, and missing information. Flag conflicts before doing the work.

   Review and correct the summary.

6. Create the First Result

   Ask the chatbot to complete the task using the curated brief.

7. Change One Context Variable

   Change the audience, source of truth, key constraint, outdated decision, example, or preference. Ask for a revised result.

8. Compare

   Identify what changed, what the chatbot inferred, what remained stable, and which result better fits the goal.

9. Maintain

   Choose what should be added, corrected, removed, isolated, retained for later, or discarded when the task ends.

Check your result

• Explain which context mattered most, what you excluded, what required verification, and whether more information improved the result or introduced noise.

• A polished result is not proof that it is accurate, complete, safe, or appropriate. Keep unresolved questions visible and verify consequential claims outside the tool.

You are done when

Save the context inventory, source hierarchy, clean brief, first and revised results, one-variable comparison, maintenance decision, and a short reflection. Done means you can explain one context decision that improved fit, one exclusion that reduced risk or noise, and one question that still needs human verification.

• The low-risk task and audience
• The source of truth and supporting sources
• The required, useful, optional, excluded, and verify decisions
• A clean context brief
• The first result and one changed condition
• A comparison of what changed
• A maintenance decision and one human verification need

Try another version

These are optional extensions, not additional completion requirements.

• Change a different context variable and compare again
• Ask a partner to sort the same context before seeing your decisions
• Create a shorter brief without removing a required constraint
• Repeat later or with another tool and record variation

Learner Map

Tracks

• AI Literacy for the Rest of Us

Modules

• Context Engineering

Help us improve this Activity

Tell us what helped, what slowed you down, or what you needed but could not find.

🎓 Give learner feedback