引導式測驗 · 中等 等級
如何使用 AI 測驗編寫程式碼文檔
檢查如何提示 AI 提供程式碼文件、根據實施驗證範例並保持生成的解釋準確。
問題 1 的 8
助理為驗證文件路徑的函數起草文件字串。開發者首先應該驗證什麼?
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。
- 助理為驗證文件路徑的函數起草文件字串。開發者首先應該驗證什麼?
- The described inputs, return value, side effects and edge cases against the code and tests
- Whether the prose uses the most technical vocabulary
- Whether the comment is longer than the function
- Whether the model sounds confident
- 自述文件說要運行人工智慧產生的設定命令。什麼是強驗證步驟?
- Assume it works because the command is familiar
- Remove all installation steps
- Run it in a clean environment and confirm it matches the project scripts and configuration
- Ask the assistant to repeat the command with more confidence
- 哪個內嵌註解最有用?
- A paraphrase of the next line with no added context
- A note explaining a non-obvious invariant or reason for the code
- A claim that the function is secure without evidence
- A list of unrelated TODOs
- 產生的 API 範例匯入儲存庫中未使用的合理套件名稱。應該發生什麼事?
- Keep it because the package name looks conventional
- Change the dependency file to match the example without review
- Delete the API documentation
- Check the import against actual dependencies and run the example
- 為什麼開發人員應該避免將機密發送給未經批准的編碼助理?
- The code or context may be processed under data-handling terms that have not been approved
- Secrets make comments harder to read
- The assistant needs every credential to understand a function
- A README cannot describe a credential safely
- A performance paragraph claims an algorithm is O(n), but no analysis or test supports it. What is the responsible edit?
- Keep it because asymptotic notation sounds precise
- Ask the model to invent a benchmark
- Remove or qualify the claim until it is verified
- Replace it with a security guarantee
- A new CLI flag changes behavior. When should its documentation be updated?
- Only after users report that the README is wrong
- With the code change and review, including a checked example
- Never, because flags are discoverable from source
- After deleting the tests
- Why might a code assistant describe an incorrect design reason even when it reads the function?
- Source code contains every team decision automatically
- Tests always explain product intent completely
- Documentation never needs human input
- The rationale or operational context may not be inferable from source code alone