What does GitHub advise when asking Copilot to build something large?
Why B is the answer
Break it into multiple simple, small tasks
Smaller steps are easier for Copilot to do well and easier for you to review.
In detail
- GitHub advises breaking large work into simple, small tasks.
- Each step is then easier to check.
- Review after each step, not only at the end.
Use case
An illustrative scenario, not a real incident.
Asked for a whole feature at once, Copilot gives a shallow result. Split into data model, API and interface, each step comes out better and is easier to check.
See it in different ways
flowchart LR A["Start general"] --> B["Add specifics"] B --> C["Give examples"] C --> D["Break up big tasks"] D --> E["Review and refine"]
mindmap
root((Prompting))
Examples
Zero-shot
One-shot
Few-shot
Structure
General, then specific
Small tasks
Target
Tests first
Context
New thread when staleflowchart LR S0["Data model"] --> S1["API"] --> S2["Interface"] --> S3["Tests"]
Key point: Break large work into small tasks.
Why the other options are wrong
- A. Put the whole request in one very long prompt
- One very long prompt is what the advice warns against.
- C. Ask for the entire result in a single word
- One word gives Copilot almost nothing to work with.
- D. Avoid describing the goal, and let it guess
- Copilot needs to know the goal.