kubernetes-data-platform

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt contains literal secret values (AWS-like keys, MinIO credentials, a GitHub token string, and an inline Fernet key generation used in a kubectl command), so following or reproducing the examples would cause the LLM to output secrets verbatim.

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). The URLs are mostly internal hosts and official sites, but the embedded credential URL (https://token:ghp_xxx@github.com) exposes a GitHub personal access token in plaintext which is a high-risk credential leak that could be abused to fetch or distribute malicious code.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill contains runtime commands that fetch and execute external content—helm pulls the Airflow chart from https://airflow.apache.org, git-sync clones https://github.com/org/airflow-dags.git (with credentials https://token:ghp_xxx@github.com), and Spark/airflow reference multiple s3a:// and s3:// URIs (e.g., s3a://data-lake/jobs/silver_etl.py, s3a://data-lake/k8s/driver-template.yaml, s3a://data-lake/k8s/executor-template.yaml, s3a://data-lake/jars/openlineage-spark_2.12-1.15.0.jar, s3a://data-lake/spark-history, s3://data-lake/airflow-logs, s3a://data-lake/warehouse) which are fetched at runtime and provide code/templates that control what is deployed/executed.