Domain 4: Apply Prompt Engineering and Context Crafting (10–15%)

← Domain 3 · Next Domain →

Exam Tip

Prompt engineering is about giving Copilot the right information to generate the right output. The exam tests your ability to recognize a well-structured prompt and know when to apply zero-shot vs. few-shot techniques.

---

Prompt Structure and Context

A high-quality Copilot prompt has four elements:

Element	Purpose	Example
Goal	What you want Copilot to do	"Write a function that..."
Context	Why or in what situation	"...for a REST API that handles user auth"
Source	What data/files to reference	"...using the User class defined in models.py"
Format/Expectation	What the output should look like	"...return a typed dict, include error handling"

Exam Tip

A comment above a function is one of the most powerful prompts in Copilot. Clear, descriptive comments produce better inline suggestions than vague ones.

---

How Context is Determined

Copilot automatically uses:

• Current file: Highest priority — the code around the cursor
• Open tabs: Other files currently open in the editor contribute context
• Language/framework: Copilot adjusts suggestions based on detected language and imports
• Comments: Inline and block comments are interpreted as instructions

To improve context:

• Keep relevant files open in editor tabs
• Write descriptive function/variable names
• Add comments describing intent, not just what the code does

---

Zero-Shot Prompting

Zero-shot prompting: asking the model to perform a task without any examples.

# Zero-shot: just describe what you want
# Parse a JSON string and return the 'email' field, or None if missing
def get_email(json_str: str) -> str | None:
    # Copilot completes here

When to use: Simple, well-defined tasks where the intent is clear from context alone.

---

Few-Shot Prompting

Few-shot prompting: providing examples before the task to guide the model.

# Few-shot: show the pattern, then ask for more
# Input: "hello world" → Output: "Hello World"
def title_case(s: str) -> str:
    return s.title()

# Input: "remove spaces" → Output: "removespaces"
def remove_spaces(s: str) -> str:
    return s.replace(" ", "")

# Now generate: Input: "reverse me" → Output: "em esrever"
def reverse_string(s: str) -> str:
    # Copilot completes here, following the established pattern

When to use: When the pattern or style is non-obvious or when you want Copilot to follow a specific convention.

---

Best Practices for Prompt Crafting

• Be specific: "Write a function that validates an email address using regex" > "Write email validation"
• Specify the output: Include expected return types, error handling style, and format
• Use personas: Start with "As a senior Python developer..." to influence code quality and style
• Reference files: Mention specific files or classes when relevant (see the models.py User class)
• Iterate: If the first suggestion is wrong, refine your prompt and try again
• Prompt files: Store reusable instructions in .github/copilot-instructions.md to establish workspace-wide conventions

---

Prompt Engineering Principles

Principle	Meaning
Clarity	Unambiguous language produces better suggestions
Specificity	Detailed prompts outperform vague ones
Examples	Showing the pattern (few-shot) reduces ambiguity
Iteration	Treat Copilot as a conversation — refine based on output
Context management	Keep relevant files open; close irrelevant ones to reduce noise

---

Chat History Usage

In Copilot Chat, previous messages in the session are included in the context window:

• Copilot uses chat history to maintain conversational continuity
• If you change topics, consider starting a new chat to avoid old context contaminating new suggestions
• Long conversations may hit context limits — earlier messages may get truncated

Domain 4 Quick Quiz

1 / 4

❓

What is the difference between zero-shot and few-shot prompting?

(Click to reveal)

💡

Zero-shot: describe the task without examples. Few-shot: provide 1-3 examples of the pattern before asking Copilot to continue it.

---

← Domain 3 · Next Domain →