Prompt file imported from gderossilive/AzSreAgentLab (.github/prompts/Demo6-GrubifyIssueTriage.prompt.md). Fill in {{AGENT_ENDPOINT}}, {{TOKEN}} before use. Copyright stays with the author.

Demo6-GrubifyIssueTriage

Goal

Run Act 3: Workflow Automation — seed sample customer issues in a GitHub repo and let the issue-triager subagent classify, label, and comment on each one.

Prerequisites

1) GitHub integration is configured

cd /workspaces/AzSreAgentLab/demos/GrubifyIncidentLab
azd env get-value GITHUB_PAT 2>/dev/null | head -c 10 && echo "... (PAT set)"
azd env get-value GITHUB_USER 2>/dev/null

2) Verify issue-triager subagent exists

AGENT_ENDPOINT=$(azd env get-value SRE_AGENT_ENDPOINT 2>/dev/null)
TOKEN=$(az account get-access-token --resource https://azuresre.dev --query accessToken -o tsv)
curl -s "{{AGENT_ENDPOINT}}/api/v2/extendedAgent/agents/issue-triager" \
  -H "Authorization: Bearer {{TOKEN}}" | python3 -c "import sys,json; d=json.load(sys.stdin); print(f'Name: {d[\"name\"]}')"

Execution

Step 1: Create sample customer issues

cd /workspaces/AzSreAgentLab/demos/GrubifyIncidentLab

# Set the target repo (owner/repo format)
export GITHUB_PAT=$(azd env get-value GITHUB_PAT 2>/dev/null)
./scripts/create-sample-issues.sh <owner/repo>

This creates 5 realistic customer-reported issues with [Customer Issue] in the title:

• App crashes when adding items to cart (api-bug / memory-leak)
• Menu page loading slowly (performance)
• Can't place an order — 500 error (api-bug)
• Feature request — add restaurant search (feature-request)
• How do I clear my cart? (question)

Step 2: Trigger the issue-triager

The issue-triager runs on a scheduled task (every 12 hours via cron). To verify it's set up:

AGENT_ENDPOINT=$(azd env get-value SRE_AGENT_ENDPOINT 2>/dev/null)
TOKEN=$(az account get-access-token --resource https://azuresre.dev --query accessToken -o tsv)
curl -s "{{AGENT_ENDPOINT}}/api/v1/scheduledtasks" \
  -H "Authorization: Bearer {{TOKEN}}" | python3 -c "
import sys,json
for t in json.load(sys.stdin):
    print(f'{t[\"name\"]} ({t[\"cronExpression\"]}) → {t.get(\"agent\",\"(none)\")} [{t.get(\"status\",\"?\")}]')
"

You can also manually trigger the triager from the SRE Agent portal at https://sre.azure.com.

Step 3: Review triaged issues

After the triager runs, check your GitHub issues. Each [Customer Issue] should have:

• Classification labels applied (e.g., api-bug, memory-leak, feature-request, question)
• A triage comment with analysis, classification rationale, and next steps

Success Criteria

• 5 sample issues created successfully in the GitHub repo
• Scheduled task triage-grubify-issues is active
• After triage runs: issues have labels applied and triage comments posted
• Classifications are reasonable (cart crash → api-bug/memory-leak, search request → feature-request, etc.)

Constraints

• Requires GITHUB_PAT with repo scope
• The GitHub MCP connector must be active
• Issue titles must contain [Customer Issue] to be picked up by the triager