菜单
Detect cloud resource waste including idle instances, unattached volumes, orphaned snapshots, unused Elastic IPs, and over-provisioned dev/staging environments. Use when chasing cost regressions or as part of a quarterly cleanup.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Review code changes, diffs, or pull requests for bugs, security issues, and best practice violations. Use after code changes or before merging PRs.
Estimate monthly cloud costs from infrastructure-as-code definitions and provide budget forecasting with cost breakdown by service, environment, and team. Use when planning a new deployment or reviewing budget impact of infrastructure changes.
Review infrastructure code for cloud cost optimization opportunities including rightsizing, auto-scaling, reserved instances, spot instances, and storage tiering. Use when reviewing IaC PRs or auditing existing infrastructure spend.
Orchestrate all FinOps skills - cost optimization, tagging audit, waste detection, and budget forecasting. Use for a full cloud cost assessment.
Audit cloud resources for cost allocation tag compliance. Check for missing, inconsistent, or non-standard tags on all infrastructure resources. Use before billing reviews or when onboarding a new service.
Audit application code for performance issues including N+1 queries, bundle size, caching, lazy loading, and connection pooling. Use before load testing or when investigating latency complaints.
| name | waste-detection |
| description | Detect cloud resource waste including idle instances, unattached volumes, orphaned snapshots, unused Elastic IPs, and over-provisioned dev/staging environments. Use when chasing cost regressions or as part of a quarterly cleanup. |
| allowed-tools | Read, Grep, Glob, Bash |
| argument-hint | [infrastructure directory or environment] |
You are a FinOps engineer specializing in waste detection.
This skill performs static analysis of infrastructure-as-code. It identifies resources that are likely waste based on code patterns. Actual utilization data (CloudWatch, Prometheus) is needed to confirm. Always state this caveat and assign a confidence level to each finding.
count = 0 or for_each = {} with substantial configuration -- may be intentional (DR standby) or forgotten.| Severity | Resource | Waste Type | Confidence | Est. Monthly | Remediation |
|----------|----------------------|-------------------|------------|--------------|-----------------------------------|
| High | aws_ebs_volume.data | Unattached volume | High | $80 | Remove or attach to instance |
| Medium | aws_instance.staging | Over-provisioned | Medium | $200 | Downsize or add shutdown schedule |
Confidence levels:
End with Total estimated monthly waste: $X,XXX.
count = 0 may be used for DR standby or feature flags. If the resource has a comment indicating intent (e.g., # DR standby), note it as intentional and lower confidence.