Skip to content

fix(12_Azure_Copilot): align lab content with real Azure Copilot behavior#345

Merged
qxsch merged 1 commit intomicrosoft:mainfrom
nekdima:pr/ch12-copilot-e2e-fixes
Apr 27, 2026
Merged

fix(12_Azure_Copilot): align lab content with real Azure Copilot behavior#345
qxsch merged 1 commit intomicrosoft:mainfrom
nekdima:pr/ch12-copilot-e2e-fixes

Conversation

@nekdima
Copy link
Copy Markdown
Contributor

@nekdima nekdima commented Apr 24, 2026

Summary

Updates the Azure Copilot MicroHack content so challenge prompts, walkthrough expectations, and deployed resources better match observed Azure Copilot behavior in a real tenant.

What changed

  • Tightens challenge prompts across the specialized Azure Copilot agents so participants get more consistent results.
  • Updates walkthrough expectations to account for realistic Copilot variability instead of relying on stale exact responses.
  • Clarifies agent-mode usage, plan-canvas workflows, follow-up prompts, and no-cluster-required AKS troubleshooting guidance.
  • Aligns Challenge 5/6 VM SKU references and cost notes with the deployed D-series VM sizes used for more deterministic agent scenarios.
  • Keeps the PR scoped to 12_Azure_Copilot content and IaC only.

Validation

  • Tested through live Azure Copilot E2E rounds in a real tenant.
  • Re-reviewed with a focused MicroHack readiness audit before opening the PR.
  • Checked the final branch is based on upstream/main with a single squashed commit.

…t E2E behavior

Round 4 of end-to-end testing against live tenants revealed consistent drift between the walkthroughs, challenge prompts, and what Copilot actually produces in the Azure portal today. This squashes five in-progress fixes into one clean commit for upstream review.

Changes:

• Challenges 1-7: tighten prompt wording so they land on reliable Copilot behaviors (grounded answers, consistent Advisor output, deterministic Resource Graph queries).

• Walkthroughs 1-7: replace stale expected answers with observed outputs (structural patterns, resource names, section headings), trim over-long model responses that never reproduce.

• Ch1 t5-2/3: reword Advisor prompts for reliability (Copilot now surfaces recommendations inline rather than deep-linking).

• Ch4/Ch5 Bicep: align iac modules with walkthrough RG names and data tier used in actual test runs.

• Setup hardening (from 18d2749): Bicep-only fragment — no setup script changes included here (those live in the fork-specific sovereign-cloud automation).

Net effect: +184 / −271 lines across 16 files (7 challenges + 7 walkthroughs + 2 Bicep modules). No behavioral regressions; tested against db467d04-dacb-474a-9dae-57d2527101f4 with 3 participants.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
@nekdima
Copy link
Copy Markdown
Contributor Author

nekdima commented Apr 24, 2026

@qxsch please review and approve merge.

@qxsch qxsch merged commit ccac193 into microsoft:main Apr 27, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants