Context Engineering Quiz

Question 1A team rewrites a request clearly, but the needed private record never reaches the model. What further work addresses the failure?

Replacing the request with a more polite sentenceExpanding the response length to create the absent factsDesigning the information supply around the requestRelying on polished language to supply the missing evidence

Question 2Retrieval returns a relevant paragraph, but assembling the input cuts off its exception. Which stage needs attention?

The filtering and packaging of retrieved evidenceTraining on examples of grammatical sentencesAuthorization checks on the final action after generationThe output validator that checks response field names

Question 3A search result closely matches a request but belongs to a retired policy. What makes it unsuitable as the governing source?

Its similarity score is too close to the requestIts section heading is shorter than the active policy’s headingIts vocabulary matches the wording used by the requesterIts effective period does not cover the current decision

Question 4Inside a retrieved document is a command to expose another person’s records. What limits the damage?

The document’s position at the end of the assembled inputA source label granting its text permission to run commandsAccess controls enforced by the application outside generationThe writer’s confident tone overriding the account permissions

Question 5Summarizing a workflow omits an unusual error condition. How should the system preserve a route back to that detail?

An identifier linking the summary to the original artifactA shorter replacement summary containing the error’s titleA confidence label attached to the compressed descriptionThe latest draft answer stored as the diagnostic record

Question 6After a repair, a test run succeeds after an earlier failure. Which working-state update fits the new evidence?

The latest observation supersedes the prior result as the current statusThe failed run and later success combine into a third invented outcomeThe earlier failure stays the current status because it began the investigationThe new success changes the original error log to show a pass

Question 7An application has room for additional tokens. Which use of that capacity is justified?

Unrelated transcript fragments chosen to increase input lengthEvidence needed to decide the task’s unresolved questionsDuplicate examples selected to occupy the remaining spaceAn obsolete plan retained because it has already been written

Question 8A model receives a description of a lookup tool. What is available before that tool is called?

A completed external action created by reading the interfaceThe requested lookup’s result inferred from its parameter namesThe live database entry already embedded in the descriptionA capability specification awaiting a requested operation

Question 9Developers change retrieval, chunking, and instructions together, then see a better outcome. What remains uncertain?

The number of components that the team changedThe contribution made by each altered partThe inclusion of chunking among the revised componentsThe existence of an improvement in the measured outcome

Question 10Diagnostic logs must help trace failures while protecting private information. Which record is the best fit?

An aggregate token total replacing the retrieval and check recordsPersonal account details copied into an unrestricted transcriptA final answer saved in place of its source referencesTraceable references and masked fields alongside check outcomes

Answer every question to see and share your score.

Context Engineering Quiz

How does it work?

What will you get?