redpanda

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt contains numerous hard-coded credentials and examples that embed secrets verbatim in commands and configs (e.g., --password in rpk commands, sasl.password in Python, jaas config), which would require the LLM to reproduce secret values directly and thus create an exfiltration risk.

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

• Potentially malicious external URL detected (high risk: 0.80). The Kubernetes instructions run "helm repo add https://charts.redpanda.com" and "helm install redpanda/redpanda", which at runtime fetch and install remote Helm chart content from https://charts.redpanda.com (executing remote templates/manifest installation), so this is a runtime external dependency that can execute remote code.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The skill includes instructions that require root/sudo on host machines (e.g., "rpk redpanda tune all" — "run as root on each broker host") which ask operators to modify OS-level tuning and system settings, so it pushes actions that can change the machine state; the other user-creation commands are application-level ACLs, not OS user accounts.