external/anthropic-cybersecurity-skills/skills/enumerating-cloud-with-cloudfox/SKILL.md
Map AWS and Azure attack paths and find exploitable misconfigurations with CloudFox.
npx skillsauth add seikaikyo/dash-skills enumerating-cloud-with-cloudfoxInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Legal Notice: This skill is for authorized cloud penetration testing and assessment only. CloudFox makes read/describe API calls against the cloud account whose credentials you supply. Run it ONLY against accounts you own or are authorized to test under a signed scope. Although CloudFox is read-only by design, the enumeration it performs is reconnaissance against a live environment and must be in scope.
CloudFox is an open-source command-line tool from Bishop Fox that helps penetration testers and red teamers gain situational awareness in unfamiliar cloud environments. Where tools like ScoutSuite focus on a defender-style configuration audit, CloudFox is built from the attacker's perspective: it answers questions like "what are the most attackable secrets, endpoints, and instances in this account, and what can the identity I just compromised actually reach?" It is read-only — it only performs Describe/List/Get style calls — and writes its findings to per-command CSV/TXT/loot files plus a combined report directory, so output can be triaged offline.
CloudFox covers AWS most deeply (30+ commands) and supports Azure. The workhorse is cloudfox aws all-checks, which runs the full battery of enumeration commands with sensible defaults: inventory, internet-reachable endpoints, EC2 instances (with IPs and instance-profile roles), iam-simulator and permissions for IAM analysis, principals, secrets from Secrets Manager/SSM, buckets, role-trusts (which identities can assume which roles — a core attack-path primitive), access-keys, route53, ecr, lambda, and more. CloudFox also emits ready-to-run command suggestions (e.g. aws s3 ls lines, aws ssm start-session lines) in its "loot" files so an operator can pivot immediately.
This skill covers installing CloudFox, authenticating to AWS and Azure, running targeted and full enumeration, interpreting the high-value outputs (role-trusts, secrets, endpoints), and feeding the results into attack-path planning. Source: github.com/BishopFox/cloudfox.
sts:AssumeRole trust relationships to plan lateral movement / privesc# Homebrew
brew install cloudfox
# Go (1.21+)
go install github.com/BishopFox/cloudfox@latest
# or download a release binary from GitHub and chmod +x
# AWS — configure a named profile and verify
aws configure --profile assess
aws sts get-caller-identity --profile assess
# Azure
az login
az account show
awscli (AWS) and/or azure-cli (Azure) installed for credential setup and follow-up| ID | Name | Use in this skill |
|----|------|-------------------|
| T1526 | Cloud Service Discovery | CloudFox enumerates the available cloud services and resources in an account |
| T1580 | Cloud Infrastructure Discovery | inventory, instances, buckets map the infrastructure footprint |
| T1087.004 | Account Discovery: Cloud Account | principals, access-keys enumerate cloud identities |
| T1069.003 | Permission Groups Discovery: Cloud Groups | permissions, iam-simulator, role-trusts reveal entitlements |
| T1538 | Cloud Service Dashboard | Aggregated situational-awareness reporting across services |
aws sts get-caller-identity --profile assess
cloudfox aws --profile assess all-checks -o ./loot
cloudfox aws --profile assess inventory
cloudfox aws --profile assess endpoints
cloudfox aws --profile assess instances
role-trusts is the key lateral-movement primitive — it shows who can assume what.
cloudfox aws --profile assess principals
cloudfox aws --profile assess permissions
cloudfox aws --profile assess role-trusts
cloudfox aws --profile assess access-keys
cloudfox aws --profile assess secrets
cloudfox aws --profile assess buckets
cloudfox aws --profile assess ecr
cloudfox aws --profile assess lambda
cloudfox aws --profile assess route53
cloudfox aws --profile assess iam-simulator
CloudFox Azure works against the subscriptions the az session can see.
cloudfox azure inventory --outdir ./azure-loot
cloudfox azure rbac
cloudfox azure storage
cloudfox azure vms
CloudFox writes per-command CSV/TXT plus a loot directory of pivot commands.
ls -R ./loot/cloudfox-output/
# Loot files contain ready-to-run follow-ups, e.g. aws s3 ls / ssm start-session lines
See scripts/agent.py to run a curated set of commands and summarize output files.
| Resource | Purpose | Link | |----------|---------|------| | CloudFox GitHub | Source, releases, full command list | https://github.com/BishopFox/cloudfox | | CloudFox docs/wiki | Per-command output explanations | https://github.com/BishopFox/cloudfox/wiki | | Bishop Fox CloudFox blog | Design and usage walkthrough | https://bishopfox.com/blog/introducing-cloudfox | | AWS CLI reference | Follow-up exploitation commands | https://docs.aws.amazon.com/cli/latest/reference/ | | Pacu | Active exploitation after enumeration | https://github.com/RhinoSecurityLabs/pacu |
CloudFox is read-only, but its enumeration is far from silent. Each command issues
many Describe*/List*/Get* API calls in a short burst, which is highly visible
to defenders:
iam:ListUsers, iam:ListRoles,
secretsmanager:ListSecrets, ec2:DescribeInstances, and sts:GetCallerIdentity
from one principal within seconds is a strong enumeration signal.Discovery:IAMUser/AnomalousBehavior and
Discovery:S3/MaliciousIPCaller can fire on this burst pattern.For an authorized assessment, document the source IP and timestamp of CloudFox runs so the blue team can correlate, and prefer running from an in-scope, attributable host.
all-checks once to populate the full output directory.role-trusts first — it reveals the assume-role graph for lateral movement.secrets and env-vars for credentials that unlock new principals.endpoints + instances to map externally reachable attack surface.| Command | Why it matters |
|---------|----------------|
| all-checks | Runs the full enumeration battery with defaults |
| role-trusts | Maps assume-role paths — core for lateral movement/privesc |
| endpoints | Surfaces internet-reachable attack surface |
| secrets | Exposes credentials in Secrets Manager / SSM |
| permissions | Lists effective IAM permissions per principal |
| instances | EC2 with IPs and attached instance-profile roles |
| access-keys | Active access keys (potential credential targets) |
cloudfox aws --helpsts get-caller-identity / az account showall-checks completed and output directory populateddevelopment
拋棄式 HTML mockup 比稿:產出 2 到 3 個設計立場不同的變體(密度 / 版式 / 強調軸,不是換色),各附取捨說明,最後給有立場的對比結論。適用:「畫個草圖」「比較 A 版 B 版」「先看方向再做」「給我看幾種做法」。要 production 元件或設計已定案時不適用。
tools
需求不明時的意圖萃取訪談:一次一題、每題附上自己的猜測、聽出「真正想要 vs 覺得應該要」,直到能預測使用者反應(約 95% 信心)才動工。適用:需求缺少對象 / 動機 / 成功標準 / 約束,或使用者點名「訪談我」「先確認一下」「我們確定嗎」。明確自足的指示、純資訊查詢、機械性操作不適用。
development
對非平凡決策啟動新鮮 context 對抗審查(找碴不背書),在修正還便宜的時候抓出錯誤方向。適用:高風險改動(production、資安敏感邏輯、不可逆操作)、不熟的程式碼、要宣稱「這樣是安全的 / 可行的」之前。機械性操作與一行修改不適用。
testing
Reference for writing and editing agent skills well — the vocabulary and principles that make a skill predictable. Consult when authoring, reviewing, or pruning a SKILL.md.