Do's and Don'ts of Working in a Space
Do keep your questions tightly scoped to the sources you attached—files, issues, pull requests, and notes—so answers stay grounded.
Don't @‑mention people or other Copilot extensions in a Space: user mentions won't notify anyone, and extensions can't be invoked from Space chats.
Do treat the Space as a focused environment for a single task or domain, and reuse its own terminology to reinforce consistency.
Do use prompting patterns that lead to runnable, verifiable outputs. Start by confirming intent, then refine with concrete constraints (formats, time ranges, file paths, or sections to consider). Ask for executable code, queries, or commands and, when helpful, request references back to the included sources for traceability.
Don't expect the Space to pull in content that isn't included—unless your environment supports repository search you've explicitly attached, Copilot won't discover outside material.
Do iterate when responses drift: tighten instructions, add one to three high‑quality examples that demonstrate "good" output, and prune noisy or irrelevant sources.
Don't let the Space sprawl beyond a single job or exceed model context limits; if you hit size warnings or degraded answers, reduce sources or split into smaller Spaces to restore precision and predictability.
Do keep context fresh and well‑ordered. Link version‑controlled files so the Space reflects your repository's default branch as it evolves, and lead with the most important sources or examples since ordering can influence responses.
Don't paste sensitive data into free‑text notes; prefer linking to files in repos (or use uploads where supported) so standard review and permissions apply.