- Home
- Skills
- DevOps & Cloud
- AWS Well-Architected Report Agent Team
More screenshots
Works with the AI Agents you already use
AWS Well-Architected Report Agent Team
Read-only AWS scan by a Claude Code agent team → four-pillar Well-Architected report, written in Traditional Chinese.
$5
AWS Well-Architected Report Agent Team
Example session with this skill installed
/report-aws 2026-06
- Read your context and instructions
- Compiled the aws well-architected report
- Generated the document
/report-aws 2026-06 完成
全程唯讀、無人值守跑完六個階段(掃描 → 四支柱並行 → 彙整 → HTML → 連結檢查 → 存檔)。
以下為遮罩示意版:帳號 ID 已遮罩,資源名稱與網域已替換為示意名稱。
- 期別:2026-06(成本窗 2026-04 ~ 2026-06,MONTHLY)
- 帳號:
1234****5678區域:us-east-1(全域資源另以 global 擷取) - 掃描日期:2026-07-24
四支柱評分
| 支柱 | 分數 | 高 | 中 | 低 |
|---|---|---|---|---|
| 安全性 Security | 2 / 5 | 5 | 6 | 2 |
| 可靠性 Reliability | 3 / 5 | 1 | 3 | 2 |
| 效能效率 Performance | 4 / 5 | 0 | 1 | 3 |
| 成本最佳化 Cost | 2 / 5 | 1 | 5 | 4 |
| 合計 | 7 | 15 | 11 |
最優先處理的五件事
關閉正式資料庫的公網暴露(SEC-03、SEC-11):app-db-prd 設為 PubliclyAccessible,所在的「private」子網實際有 0.0.0.0/0 → IGW 路由(由 network-facts.py 確定性算出)。
2. 封鎖 S3 資料桶公開存取(SEC-04):3 個 bucket 的 Public Access Block 四項全關。
3. root 與所有使用者啟用 MFA、開啟 CloudTrail(SEC-01、SEC-02):帳號零稽核軌跡。
4. 建立監控告警(REL-01):全帳號零 CloudWatch 告警、零 SNS 主題。
5. 整併重複的 ALB 並清理孤兒資源(COST-01、COST-07):6→2 台 ALB +刪除 2 個孤兒 WAF ACL,約省 $76/月。
成本
已盤點結構性節省約
$126/月(約佔月均帳單 36%);另有 RDS RI/Savings Plan 空間待 Cost Explorer 核實。覆蓋率自檢標出 cloudformation/glue/servicediscovery 有計費但未掃明細。
連結檢查
每項建議都附 AWS 官方文件連結,由 check-links.sh 驗證;若有失效連結會在此列出並提示更新 references/ 目錄檔。
產出檔
report/aws-report.html(確定性 HTML,版型逐月固定)report/AWS架構報告.md、report/report-data.json- 已存檔至
archive/2026-06/
下一步:對外分享請執行 node .claude/skills/report-aws/scripts/build-report.js --masked(殘留帳號 ID、金鑰、ARN、私有 IP、Email 即拒絕輸出);需要架構圖可再跑 /aws-diagram。

aws-report.html
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Output language: Traditional Chinese (AWS service names kept in English).
Type /report-aws in Claude Code and a team of six agents scans your real AWS account, analyzes it against the four
AWS Well-Architected pillars, and writes a Markdown + HTML architecture report — unattended, from start to finish.
- Read-only guarantee — only
describe-*/list-*/get-*APIs. Designed for an IAM identity withReadOnlyAccess, backed by a deny list for write verbs and a PreToolUse hook that blocksawscommands with shell expansions. - Evidence-driven — every finding cites the actual scan data. Cross-file facts (e.g. whether RDS sits in a subnet routed to an IGW) are computed by scripts, not guessed by the LLM. A coverage self-check flags services you are billed for but that were not scanned.
- Four pillars in parallel — Security, Reliability, Performance Efficiency, and Cost Optimization agents run concurrently; a report-writer agent consolidates them. Every recommendation links to official AWS documentation, and links are validated by script.
- Deterministic HTML — the HTML report is rendered from a frozen template + structured data without an LLM, so layout and sections are identical every month and reports are comparable over time. Swappable color themes.
- Masked report for sharing — reports are unmasked by default for internal use;
--maskedbuilds an external-sharing version and refuses to output it if any 12-digit account ID, access key, ARN, private IP, or email address slips through. - Optional draw.io architecture diagram —
/aws-diagramturns the same scan data into a multi-page, subnet-level.drawiotopology, deterministically generated with count assertions. - Period archive — each report is archived per period for cross-period regression comparison.
Requirements
- Claude Code
- AWS CLI v2,
jq,python3, Node.js - AWS credentials — an IAM identity with
ReadOnlyAccessis strongly recommended - Install the whole project (
.claude/+CLAUDE.md) at the project root — this is a multi-agent project, not a single skill. SeeREADME.en.md.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.
Reports are generated in Traditional Chinese only. Built for Claude Code (relies on subagents, settings.json permissions and a PreToolUse hook); other agents may not support these. Requires AWS CLI credentials, ideally ReadOnlyAccess. Diagrams are script-generated; manual edits to output files are overwritten on rerun.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 26 days ago
- Accounts
Amazon
- Software
- Claude CodeAWS CLI v2jq
Python 3
Node.js
Git
Permissions
Allowed Hosts
Shell: runs read-only AWS CLI calls (describe/list/get only) plus local jq/python3/node scripts. Network: AWS APIs and an AWS docs link check. Files: writes only to data/, findings/, report/, archive/ inside the project. Use AWS credentials with the ReadOnlyAccess policy.