Agent Automation Controls

What This Demonstrates

GitHub Issues can now show why an agent suggests changes like labels, issue type, assignee, or status/closure. You can review each suggestion before it applies, accept or reject it, and keep a history of what was approved or declined.

Prerequisites

• A repository with Issues enabled
• Access to create issues in codecurrent-sandbox/octo-eshop-demo
• (Optional) Existing issue automation rules or agent-driven triage enabled

Prompts

Prompt A — Trigger Auto-Labeling

Create an issue for adding Stripe payment integration to payment-service. Include product-service catalog metadata needed for payment line items, API changes using the standard success/data/error/meta response format, required Helm values for deployment to Azure AKS, and Terraform updates for Key Vault secrets.

What to look for:

• Suggested labels with an explanation of why each label is proposed
• Review controls to accept or reject each suggested change
• An activity/history trail showing accepted or rejected agent actions

Prompt B — Trigger Assignment Suggestions

Create an issue to refactor the cart-service Redis client for better retry handling and timeout behavior. Include impact on checkout performance, Redis connection pooling, and any required Helm and AKS runtime configuration updates.

What to look for:

• Suggested assignee(s) with transparent rationale
• Ability to approve/reject assignment before it is applied
• Evidence that manual control is preserved even when automation suggests owners

Prompt C — Trigger Type/Status Automation

Create a bug issue: checkout fails in order-service when user-service JWT refresh expires during order confirmation, returning inconsistent error payloads instead of success/data/error/meta. Include PostgreSQL transaction context and cross-service impact.

What to look for:

• Suggested issue type/status (for example bug or urgent triage state) with reasons
• Review-before-apply controls for type/status updates
• Logged record of which automation suggestions were accepted vs rejected

Follow-Up Prompts

• “Explain why each suggested issue change was proposed in plain language.”
• “Apply only labeling suggestions and reject assignment/type changes for now.”
• “Show me the history of accepted and rejected agent-driven issue actions on this item.”