2.3.4. Spaces, Spark, PR Summaries, and Instructions Files
💡 First Principle: This cluster is about shaping what Copilot knows and produces: Spaces and instructions files curate context and standards, PR summaries condense changes for reviewers, and Spark turns natural language into a working app. Each one steers output toward your team's intent.
- Copilot Spaces — bundles relevant context (code, docs, instructions) into a reusable "space" Copilot can draw on, so it answers grounded in the material you've curated rather than guessing.
- GitHub Spark — a natural-language app builder: describe an app in plain English and Spark generates working code with a live preview; opening a Spark project in a Codespace gives full Agent Mode for refinement. Spark is available on Pro+ and Enterprise.
- Pull request summaries — AI-generated overviews of what a PR changes, which files are affected, and what reviewers should focus on. A GitHub.com surface that speeds review.
- Instructions files — repository- or organization-level guidelines (for example, a
copilot-instructions.md) that Copilot follows so suggestions and agent output match your conventions: naming, error handling, testing frameworks, and review standards.
A scenario: a team's Copilot suggestions keep ignoring their error-handling convention. The fix isn't to nag developers — it's an instructions file that encodes the convention so Copilot's output conforms automatically.
📝 Currency note: Spaces and Spark are newer capabilities whose availability and tier gating have shifted through 2026. The exam tests their purpose — curated context (Spaces), NL app building (Spark), reviewer summaries (PR summaries), encoded standards (instructions files) — more than precise tier tables.
⚠️ Exam Trap: Instructions files customize behavior and standards; they are not the same as content exclusions (which remove files from context) or Spaces (which curate context to draw on). Match the mechanism to the goal: standards → instructions files; protect files → exclusions; bundle context → Spaces.
Reflection Question: A team wants every Copilot suggestion to follow their naming and testing conventions without manual reminders. Which mechanism encodes that, and how does it differ from a content exclusion?
