kubernetes

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

• Third-party content exposure detected (high risk: 0.80). This skill instructs the agent to fetch and apply public, third‑party resources (e.g., kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/... and kubectl apply -f https://raw.githubusercontent.com/stakater/Reloader/..., plus init-container wget of models from external URLs like https://models.example.com), which are untrusted public content the agent would pull and act on as part of its workflow, exposing it to indirect prompt‑injection via those external artifacts.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes initContainer runtime wget commands that download model binaries from https://models.example.com/sentiment-v2.bin (and https://models.example.com/model.bin), which are required at startup and directly control the agent's execution/behavior.