helm

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs retrieving Kubernetes secrets and then embedding the raw password into commands (e.g., exporting the secret and using --set auth.postgresPassword="$POSTGRES_PASSWORD" and kubectl create secret --docker-password=), which requires the LLM/user to handle and reproduce secret values verbatim.

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

• Third-party content exposure detected (high risk: 0.85). Runtime path: the workflow runs helm repo add bitnami https://charts.bitnami.com/bitnami / helm repo update and then helm search repo / helm show values/chart / helm install/upgrade, which fetches chart index and chart content from an external third-party repository (public web content authored by Bitnami) and ingests it as readable chart metadata/templates into the Helm/LLM execution context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly runs helm/docker/kubectl commands that fetch and run external charts and container images at runtime (e.g., helm repo add https://charts.bitnami.com/bitnami and pulling images like rg.fr-par.scw.cloud/chlabs-io/pgvector:17.3.0-debian-12-r0 and docker.io/bitnamilegacy/postgresql:17.3.0-debian-12-r0), which causes remote code/content to be executed as a required dependency.