Skill

securing-kubernetes-on-cloud

skill:mukul975/anthropic-cybersecurity-skills#securing-kubernetes-on-cloud

Hardens managed Kubernetes clusters on EKS, AKS, and GKE by implementing Pod Security Standards, network policies, workload identity (IRSA for EKS, Workload Identity for GKE, Managed Identities for AKS), RBAC scoping, image admission controls, and runtime security monitoring. Use when deploying a new managed Kubernetes cluster with security requirements or hardening an existing EKS, AKS, or GKE cluster after an audit or pentest finding.

SkillsMP

Verdict

Suspicious

Risky behavior the analysis could not rule out.

Analyzed Oct 27, 2026, 22:04 UTC at 54a7988

findings

medium

highest severity

threat classes

4,532

files scanned

93/93

checks satisfied

package advisories

Why this verdict

suspicious: finding skill:skills/securing-kubernetes-on-cloud:skills/securing-kubernetes-on-cloud/SKILL.md:314:SUPPLY

Security findings 1

Supply chain

mediumContextual riskmedium confidence

The skill instructs the agent to apply a Kubernetes manifest fetched from an unpinned mutable URL (raw.githubusercontent.com, main branch), so the cluster executes whatever content sits at that reference at runtime instead of pinned, reviewed content.

skills/securing-kubernetes-on-cloud/SKILL.md:314

kubectl apply -f https://raw.githubusercontent.com/aquasecurity/kube-bench/main/job-eks.yaml

vector:instruction

Counter-argument considered

The skill body concretely instructs the agent to run `kubectl apply -f` against an unpinned, mutable main-branch URL, so the cluster executes whatever content sits at that reference at runtime; this is not a pinned official-registry install (no digest/tag pin), no declared claim covers fetching external executable manifests, and the skill's own digest-pinning policies (SRC:7da53b9da5f50132) apply only to pod images and do not mitigate this fetch. The vendor-documentation status of the command does not neutralize the mutable-reference supply-chain risk.

Open questions 1

What the analysis could not settle from the code alone.

Contradiction leads C4, C5, and C6 cite D9 (a Pod Security Context field documentation table in references/api-reference.md) as a mandatory file_write safety claim broken by agent.py writing its audit report; the quoted table text contains no file-write restriction, so these contradictions could not be substantiated from the shown evidence — confirm whether D9/D10/D11 were misextracted from a reference table.