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 gave full Agent Mode for refinement. Spark was offered on Pro+ and Enterprise, but GitHub closed it on github.com in August 2026: no new users or apps from August 4, and access ended August 31 (apps already deployed keep running). GitHub now points builders to Copilot in VS Code, Copilot CLI and the GitHub Copilot app. The GH-300 objectives still name Spark, so know what it does.
• 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; Spark was closed on github.com in August 2026 even though the objectives still list it. The exam tests their purpose — curated context (Spaces), NL app building (Spark), reviewer summaries (PR summaries), encoded standards (instructions files, which the objectives call customizable review standards when applied to code review) — 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?