PRACTICAL RESOURCES
Prompts, sample files and small runnable examples from the articles. Each guide supplies the context, instructions, expected result and a condition to change. No signup required.
Prompt
Use a small, fictional evidence pack to check whether an explanation follows from the records.
Try it with: Copy the complete prompt into a new chat, or work through it yourself. Everything needed is included. This exercise uses supplied records; it does not connect to a live system or configure MCP, retrieval or memory.
Read the instructions and expected resultWorksheet
Pick one requirement that changed recently. Give a colleague its Jira story and ask them to find the reason, decision owner, impacted work and validation evidence.
Try it with: Use a requirement you are allowed to review, or one of the fictional scenarios above. Copy the worksheet, fill in the records you can locate and mark missing evidence as pending. Ask your colleague to follow the links without an extra verbal explanation.
Read the instructions and expected resultQuick check
Identify the parts, responsibilities and connections in a fictional shared task application.
Try it with: Read the small brief below and answer in three sentences, on your own or with a colleague.
Read the instructions and expected resultPrompt
Use the article’s hypothetical supplier and a delivery-change request whose response never arrived. Compare the proposed customer state and recovery process with the business rules.
Try it with: Open a new chat in your preferred AI assistant and copy the complete prompt. Its fictional fixture is included; no live order system or external data is needed. Compare the response with the expected results, then inspect the staff-change variation.
Read the instructions and expected resultPrompts + visual guide
Decide what the evidence supports, identify the missing information and revise the review when a new document arrives.
Try it with: A new chat in the model of your choice. The article includes fictional supplier files, a policy, two prompts and expected review statuses.
Read the instructions and expected resultPrompts
Use the same four-location business and the same launch constraint. First plan the staged migration; then add customer credits that must work across locations.
Try it with: Open a new chat with Astra or another reasoning model. Run the first brief, read the decision record, then send the change request in the same conversation. These records are fictional and self-contained.
Read the instructions and expected resultCode
A timeout leaves the payment outcome unknown. This small state machine makes the next action explicit and keeps a repeated event from changing the case twice.
Try it with: Download the example and run node exception-transitions.mjs with Node.js. It uses invented records and calls no payment service.
Read the instructions and expected resultWorksheet
Use a handoff review to find the responsibility hidden inside an arrow. The useful result is an agreement about the work, including what happens when the usual person is away.
Try it with: Use this as an interview worksheet or paste it into a chat to analyse the fictional notes. Validate the result with the people who own the real process before applying it.
Read the instructions and expected resultData + prompt
A table alone will not explain why somebody highlighted a row. Use the records and their accompanying notes to separate an observed workaround from an agreed business rule.
Try it with: Paste both the sample and the review prompt into a new chat. You can also save the CSV in a spreadsheet. All records are fictional; the colour meanings are supplied separately because CSV does not preserve formatting.
Read the instructions and expected resultCommands
The layers in the article leave different evidence. These read-only commands show either startup timing or the operating system’s recorded boot time; those are different measurements.
Try it with: Choose the block for your operating system and run it in a terminal or PowerShell. No administrator elevation or firmware change is part of this exercise. Availability and recorded fields vary by machine.
Read the instructions and expected resultCode + prompt
The Business Operations Platform in this article can be represented as one maintained model. This example generates its context and container views without drawing the same relationship twice.
Try it with: Download the model as workspace.dsl. Open Structurizr’s DSL editor, paste it and render the workspace, or use the local workflow linked below. All systems in this model are fictional.
Read the instructions and expected resultFile + prompt
Give the agent a screen with a defined job. This reference turns the supplier-review example into a small implementation task while leaving the existing project’s visual identity intact.
Try it with: Add the downloaded DESIGN.md to a test project you can edit, then give your coding agent the prompt below. This is a fictional screen; use a local preview for the exercise.
Read the instructions and expected resultWorksheet
Keep the experiment smaller than the problem. This worksheet makes the benefit and the cost visible before a small change becomes a permanent dependency.
Try it with: Complete it yourself or paste it into a chat. The baseline below is invented; replace it with observations before using the worksheet for a real decision.
Read the instructions and expected resultPrompt
Start with the disagreement in the article. The aim is to preserve each department’s facts while exposing the decision that crosses their boundaries.
Try it with: Paste the brief into a chat, then review the proposed glossary with the three roles in mind. The notes describe a fictional company.
Read the instructions and expected resultPrompt + schema
The useful output is an evidence-backed record that a reviewer can use. This exercise makes missing information and unreadable documents explicit without granting payment authority.
Try it with: Paste the prompt into a new chat. Compare the three returned records with the expected outcomes below; the invoices are invented.
Read the instructions and expected resultPrompt
Before choosing screens or services, identify the commitments the software has to support. This uses the client-file request from the article.
Try it with: Run the prompt in a new chat and challenge any proposed capability that has no supporting need in the notes. The example is fictional.
Read the instructions and expected resultThe separate Northstar Workshop demo lets you change a parts delay and compare a revised schedule. It is a deterministic repair-scheduling exercise, with fictional jobs.
Open the workshop demo