Lab 091: Copilot Cloud Agent and Code Review: From Issue to PR¶
Free-first design
Complete the guided exercise with Copilot Free or the included simulation. Paid features are optional and explicitly labeled.
AI Credit impact
Low for guided work; medium/high for agentic or repeated runs. Prefer automatic model selection, plan before acting, keep context focused, and avoid unnecessary parallel sessions.
What You'll Learn¶
- Trace the issue-to-PR cloud-agent lifecycle
- Review firewall, permissions, and repository instructions
- Evaluate Copilot review comments
- Estimate AI Credits and Actions-minute impact
GH-300 alignment¶
Use Copilot features (25–30%) · Data and architecture (10–15%). This mapping follows the GH-300 curriculum snapshot dated 2026-08-07. The official outline may retain older product wording such as “Copilot Edits”; this lab teaches the current product experience and calls out that difference.
Concept Map¶
sequenceDiagram
participant I as Issue
participant A as Cloud agent
participant G as GitHub Actions
participant P as Pull request
participant R as Human + Copilot review
I->>A: acceptance criteria
A->>A: plan
A->>G: commits + checks
G-->>A: evidence
A->>P: open PR
P->>R: review findings
R-->>P: approve, request changes, or close
Hands-on Exercise¶
Step 1: Free simulation: triage the issue¶
Use the supplied scenario below. Convert a vague issue into acceptance criteria, identify repository instructions, and choose a least-privileged environment.
Step 2: Free simulation: inspect the session¶
Review the synthetic timeline: plan, commits, checks, firewall requests, AI Credit band, and review findings. Decide what a human must approve.
Step 3: Free simulation: review the PR¶
Classify each Copilot comment as actionable, uncertain, or incorrect. Code review is advisory and does not replace required human approval by default.
Step 4: Optional paid path¶
If your organization enables Copilot cloud agent, assign the issue, monitor the live session, review the generated PR, request Copilot review, and record actual AI Credits and Actions minutes.
Complete Free Simulation¶
Use this synthetic session record; no Copilot license or organization is required.
| Time | Event | Evidence | Human decision |
|---|---|---|---|
| 09:00 | Issue assigned | Acceptance criteria mention tests and no new dependency | Accept scope |
| 09:03 | Plan proposed | Changes API, tests, and docs | Reject an unrelated database migration |
| 09:08 | Firewall request | Requests pypi.org for an unnecessary package |
Deny; use the standard library |
| 09:15 | Commit 1 | Implements endpoint, two tests pass | Continue |
| 09:22 | Commit 2 | Adds validation, all six tests pass | Inspect edge cases |
| 09:28 | PR opened | AI Credit band: medium; Actions: 7 minutes | Request review |
| 09:30 | Copilot review | One actionable null check, one speculative performance note | Fix the null check; document why the other is deferred |
Write a short review containing:
- The least-privileged permissions and network allowlist.
- The rejected scope and why it was rejected.
- Which review comment you would apply.
- A cost record: medium AI Credits, 7 Actions minutes, and the optimization you would try next.
Supporting kit¶
The site build generates stable PDF and ZIP downloads. The ZIP contains the lab guide and every supporting file declared for this lab.