majiayu000 Analyze Production Environment
Perform read-only production diagnosis first. Do not mutate production without explicit approval.
1. Establish Scope
Capture:
- Problem statement
- Affected service or namespace
- Environment or cluster
- Approximate start time
- Incident severity if known
2. Detect Platform
Read root AGENTS.md, deployment files, and cloud config to determine whether the target is Azure, GCP, or AWS.
Load cloud-platforms for provider-specific commands.
3. Collect Read-Only Evidence
Gather the smallest useful snapshot:
- Cluster context and node health
- Pod and deployment status
- Logs for failing workloads
- Resource pressure signals
- Ingress, service, and endpoint status
- Managed database health
- Recent warning events
4. Analyze
Look for:
- Restart loops
- OOM or probe failures
- Dependency connectivity failures
- Capacity saturation
- Deployment regressions
- Database bottlenecks
Tie every conclusion to evidence.
5. Diagnose
Return:
- Environment summary
- Findings
- Suspected root cause
- Blast radius
- Recommended mitigation
- Recommended durable fix
If a production mutation is needed, present it separately and wait for approval.
Integration
- Companion skill:
cloud-platforms - Common follow-up:
bugfix