eks-upgrade-check

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.65). The skill ingests outsider-authored free text from the cluster itself: it scans Kubernetes resources (e.g., add-on discovery via labels/images and “Unidentified Workloads” evidence) and then includes that extracted text in the generated report/LLM context; this text originates from workloads created by parties other than the operating user.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly performs live webFetch/remote_web_search during runtime to retrieve upstream compatibility and documentation pages (for example: https://karpenter.sh/docs/upgrading/compatibility/ and the compatibility_url entries in data/oss_addon_registry.json such as https://istio.io/latest/docs/releases/), and those fetched pages are required at runtime and directly drive the skill's compatibility verdicts and report text—i.e., external content controls the agent's prompts/output.