| name | idle-resource-scan |
| name-en | Idle Resource Scan |
| name-zh | 闲置资源扫描 |
| description-en | Scan an AWS account for idle, unused, or oversized resources wasting money — stopped-but-billing EBS volumes, unattached Elastic IPs, idle EC2/RDS, old snapshots, and low-utilization instances. Produces a prioritized cleanup list with estimated monthly savings per item. |
| description-zh | 扫描 AWS 账号中闲置、未使用或规格过大的浪费资源——已停止但仍在计费的 EBS 卷、未关联的弹性 IP、闲置 EC2/RDS、过期快照及低利用率实例。产出按优先级排序的清理清单，并预估每一项的每月可省金额。 |
| description | Scan an AWS account for idle, unused, or oversized resources that are wasting money — stopped-but-still-billing EBS volumes, unattached Elastic IPs, idle EC2 instances, idle RDS databases, old snapshots, and low-utilization instances. Use when the user asks to find waste, reduce cost, clean up unused resources, right-size, or find idle resources. Produces a prioritized cleanup list with estimated monthly savings per item. |
Idle & Wasted Resource Scan
You are performing a read-only waste sweep. You only describe resources and recommend
actions — you never stop, delete, or modify anything. Every finding must be backed by a
resource returned from a tool call.
When to use
- The user asks to "find waste", "reduce cost", "clean up unused resources", "right-size",
or "what can I turn off".
- A cost review surfaced spend that isn't tied to active workloads.
What to check (each is a separate finding class)
- Unattached EBS volumes — volumes in
available state still bill. List them with size
and type; estimate monthly cost from size × per-GB price.
- Unassociated Elastic IPs — an EIP not attached to a running instance bills hourly.
- Stopped EC2 instances with attached EBS — the instance is free but its volumes and any
provisioned IOPS keep billing. Flag long-stopped instances.
- Idle running EC2 — instances with sustained low CPU / near-zero network over the
trailing 1-2 weeks. Use metrics to justify "idle"; never call something idle without data.
- Idle RDS — databases with near-zero connections and low CPU over the trailing window.
- Old / orphaned snapshots — EBS/RDS snapshots far older than any sensible retention,
especially those whose source volume no longer exists.
For anything the resource-specific tools can't see, fall back to the general read-only AWS
query capability to describe/list the resource — still read-only.
Report format
Group findings by class, most-savings first:
| Resource | Type | Why it's waste | Est. monthly cost | Suggested action |
|---|
Close with a total estimated monthly savings (sum of items), a confidence note
(idle judgments are based on the metric window; a spike outside the window could change the
call), and an offer to save the full report.
Guardrails
- Read-only. Suggest actions; do not take them. Deletion/stop is always the customer's call.
- "Idle" MUST be backed by a metric or state from a tool — never assert idleness by name alone.
- Cost estimates are approximate (list price × observed size/hours); label them as estimates.
- Warn before recommending deletion of any volume/snapshot that could be a backup.
