2.3.5. Copilot Chat: Limits, Commands, and Prompt File Reuse

💡 First Principle: Copilot Chat is bounded by its context window and surface and steered by commands and reusable prompts. Knowing its limits (what it can't see, where it runs) and its controls (slash commands, prompt files) is what makes Chat reliable rather than hit-or-miss.

Chat is available in supported IDEs, on GitHub.com, in GitHub Mobile, and in Windows Terminal — but availability and depth vary by surface and plan (org-aware chat over an indexed codebase is an Enterprise capability on GitHub.com). Practical controls the exam expects you to know:

• Slash commands — shortcuts like /explain, /tests, /fix that scope a request to a common intent, producing more consistent responses than free-form phrasing.
• Feedback — you can rate responses; this signal helps improve the experience and is part of operating Copilot.
• Limits — Chat reasons only over the context it's given, within the model's context window; it doesn't magically know your whole repo unless that context is supplied (or indexed, on Enterprise).
• Prompt file reuse — reusable prompt files (Markdown in your workspace) capture a well-crafted prompt so the whole team gets consistent responses for recurring tasks, rather than re-inventing the wording each time.

💡 Key Point: Prompt files are to Chat what instructions files are to overall behavior — both encode something good so it can be reused. Prompt files capture a specific prompt; instructions files capture broad standards.

⚠️ Exam Trap: Chat's apparent knowledge is bounded by context. A question implying Chat "knows everything in the repository automatically" ignores the context window and the fact that deep codebase indexing is an Enterprise feature.

Reflection Question: Your team keeps re-typing the same elaborate prompt to generate consistent test scaffolding. Which Chat capability removes that repetition, and how does it relate to instructions files?