引導式測驗 · 中等 等級
如何向 AI 提示以取得更好的程式碼測驗
測試編碼提示的特異性、範例、相關上下文、任務分解、測試設計以及產生的變更的審查。
問題 1 的 8
What should a clear code-generation request state beyond the broad goal?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- What should a clear code-generation request state beyond the broad goal?
- A preferred language, framework, and test runner without describing the change
- Relevant context, constraints, inputs and outputs, and expected behavior
- Every neighboring module even when it does not affect the requested behavior
- A general goal with no acceptance criteria or interface details
- Why are concrete input-output examples useful in a coding prompt?
- They always reduce generated code size
- They guarantee the code is secure
- They remove the need for tests
- They clarify expected behavior and edge cases
- To reproduce a particular failing test, what context should a developer include with relevant code?
- The failing input, expected result, observed error, and relevant function or API context
- An expected result without the input that produced the failure
- Only the test name and programming language, without the actual input or output
- The full repository tree and stack trace, without identifying which test fails
- Which sequence makes a multi-part implementation request easiest to verify as work proceeds?
- Split the code by file, but defer tests and integration review until every part is generated
- Have the assistant write separate components and infer success from its completion summary
- Split it into small, testable tasks, review each result, then integrate the changes
- Ask for an outline, then request all implementation at once and review only the final output
- What does asking for tests alongside implementation help clarify?
- A test name without the inputs or expected result
- A code-style preference unrelated to observable behavior
- A list of implementation steps without saying what each step should produce
- Expected inputs, outputs, and behavior that may be ambiguous
- Why should a developer name relevant library or framework versions?
- It prevents the need to review dependency compatibility
- Available APIs can vary, and suggestions should be checked against the installed version
- It guarantees that code generated for the version has no defects
- It ensures the assistant will use only APIs introduced after that version
- Before relying on a generated code change, which review sequence is strongest?
- Review the diff, run relevant tests and checks, then assess security and compatibility
- Review formatting and generated tests, then merge without running them
- Run unit tests but skip checking interfaces and security-sensitive behavior
- Ask the assistant to confirm its code is correct, then accept its test summary
- If an assistant’s first implementation is wrong, what is a useful next prompt?
- Ask for a different implementation without identifying what failed
- Remove the failing test so the implementation can proceed
- Describe the observed discrepancy and provide a concrete failing case
- Repeat the original request with the same context and no new evidence