Start with the failed step

A red workflow tells you little. Read the failed step, its output, exit code, workflow, and recent change. Then classify the failure.

Give Copilot relevant evidence

The CI job failed in [step]. The workflow is #file:.github/workflows/ci.yml.
This change added [change].

[failing-step output]

Identify the root cause, cite evidence, and propose the smallest repair.

Do not send an entire log, secrets, or unrelated output.

Review the remediation

Use Assign to Copilot only when the customer environment supports it and policy permits it.

1. Confirm the finding and affected code.
2. Limit work to the finding and required tests.
3. Review scope, dependencies, behavior, and sensitive data.
4. Run focused tests and approved scans.
5. Require human approval.

Improve feedback without weakening controls

• Cache npm packages from the lockfile.
• Run independent lint and test jobs in parallel.
• Use shallow checkout when full history is unnecessary.
• Deploy only from pushes to main.

Branch protection and human review decide whether a finding blocks a merge.

Customer controls and fallback

Confirm the repository, tools, permissions, data boundary, reviewer, meter, and stop condition. If remote scans or agent routes are unavailable, use the supplied application and findings locally. The review standard stays the same.

Demonstration and lab

1. Show a failing matrix job and the relevant log output.
2. Compare the failed matrix legs with the workflow.
3. Diagnose the Node version, Windows path, and CI=true issues.
4. Review the minimal patch and compare it with the solution.
5. Review a prepared CodeQL remediation without merging it.