Quiz sur les outils de codage de l'IA

Les réponses et les explications apparaissent au fur et à mesure que vous répondez au quiz.

1. A coding assistant proposes an inline completion without editing files. Which tool category best describes that action?
   • An autonomous agent that has changed the repository
   • A production deployment system
   • A test suite that has validated behavior
   • A suggestion tool that proposes text for a developer to review
2. An agent can run shell commands and edit files. Which safeguard follows from the guide?
   • Give it unrestricted access because it can plan
   • Set clear boundaries and inspect observable actions before consequential changes
   • Masquer la sortie de la commande pour réduire la distraction
   • Let it publish changes without review
3. Why does providing repository conventions and relevant files help a coding tool?
   • It guarantees the code compiles
   • It expands the model’s training data permanently
   • It gives the tool project-specific context it otherwise may not see
   • It removes the need to review changes
4. A code change compiles, but the diff overwrites a user’s existing work. Which check would have helped?
   • Inspect the diff and preserve existing changes before applying edits
   • Increase model temperature
   • Ask for more generated lines
   • Skip tests to reduce runtime
5. How should a team measure whether a coding assistant improves delivery?
   • Count generated lines alone
   • Measure only first-draft speed
   • Use the assistant’s confidence score
   • Include review time, regressions, maintainability and final quality
6. Why keep API keys out of prompts sent to a coding service?
   • The key will make code compile too quickly
   • API keys cannot be stored as text
   • Prompts or generated files may expose sensitive information outside intended control
   • The tool cannot read a prompt
7. A coding model suggests an unfamiliar library function. What is the most reliable next step?
   • Assume the function exists because it sounds plausible
   • Check the official documentation for the project’s library version
   • Copy it into production immediately
   • Ask the same model to repeat the answer
8. Which approach best evaluates a coding tool on a new repository?
   • Test representative tasks against local conventions and current framework versions
   • Use a generic programming benchmark only
   • Measure only how quickly it responds
   • Let it change production code without observation