More screenshots
Works with the AI Agents you already use
Cloud and Kubernetes Security Assessment
Assess cloud, container, and Kubernetes security on authorized targets: start from identity, map what is exposed, trace blast radius, and report the top three findings with fixes and evidence.
$5
Cloud and Kubernetes Security Assessment
Example session with this skill installed
Assess our staging cluster and the payments namespace. Read-only, no production touch. What are the worst findings?
- Read your context and instructions
- Compiled the cloud and kubernetes
IDENTITY: The workload service account in payments is bound to a cluster-scoped admin role — the gap between what the SA was meant for (deploying to one namespace) and what the bindings allow (everything) is the first finding. A second identity, the CI deploy bot, can assume a cloud role with pass-role permission plus compute rights.
EXPOSURE: One storage bucket is publicly readable and holds database backups. The metadata endpoint is reachable from workloads at version one. The API server is correctly restricted.
BLAST RADIUS: The cluster-admin binding is total: any pod escape in payments becomes total cluster compromise. The pass-role chain is wide: the deploy bot can pick its own permissions. The public bucket is wide: backups are downloadable by anyone. The metadata endpoint is local-to-wide depending on what the workload's role can reach.
ESCAPE PATHS: Two workloads run privileged (deployable by anyone with deploy rights — missing admission controls), one mounts the container runtime socket read-write.
REPORT — TOP THREE BY BLAST RADIUS: (1) Cluster-admin binding on the workload SA — fix: replace with a namespace-scoped role allowing deploy and read in payments only. (2) Privileged workloads plus missing admission controls — fix: deny privileged pods at admission, remove the runtime-socket mount. (3) Publicly readable backup bucket — fix: remove public access, scope the bucket policy to the backup role. Each carries the exact resource, the attack path in plain steps, and the command output that proves it.
Connects securely to your tools. The creator never sees your data.
About this skill
Cloud breaches are identity breaches wearing an infrastructure costume — nearly every finding traces back to who an identity is, what it can reach, and what falls if it is compromised. This skill starts there, not with a port scan. Step zero is authorization: written scope naming the accounts, clusters, and namespaces, and rules that forbid destructive actions, data exfiltration, and unapproved production touch. Ambiguity is not authorization.
Then it establishes identity (cloud principals and role-assumption chains, Kubernetes service accounts and their bindings, container users and capabilities), maps the exposed surface (storage, network, metadata services, API exposure), and traces blast radius per finding — graded total, wide, local, or minimal, because a medium misconfiguration that leads to cluster admin outranks a high CVE on an isolated workload. Container escape paths are checked in likelihood order: privileged containers, host-path mounts, dangerous capabilities, kernel exposure. Every finding carries the exact resource, the attack path in plain steps, the blast radius, a specific fix (never "apply least privilege"), and command output as evidence. The report leads with the top three by blast radius. A worked example assesses a staging cluster and its payments namespace, finding a cluster-admin workload SA, privileged pods, and a public backup bucket.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Security.
- Without written authorization naming the accounts, clusters, and namespaces, the assessment does not start.
- Some findings need a write to prove; when the rules forbid it, the finding is reported as capability-confirmed by API query, exploit not executed per scope.
- A finding with no clean fix is reported with its blast radius and compensating control, not hidden.
- Checklist severity never outranks blast radius in the report ordering.
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 3 days ago