azure-hdinsight-migration-esp-to-non-esp

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow includes executing curl against the cluster’s Ranger REST endpoints (e.g., .../ranger/service/public/v2/api/policy and .../ranger/service/assets/accessAudit?...) and exporting the returned JSON into files that are then processed/printed, meaning outsider-authored free text (Ranger policy/audit content created by other admins/users) can be ingested into the agent context via the LLM-visible outputs.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill instructs the agent to run privileged commands (sudo) on cluster head nodes, access sensitive system files/keytabs, and perform destructive/cloud state-changing operations (create/delete HDInsight clusters, network/security resources, role assignments), which directly modify the machine/cloud state and can compromise security if executed by an agent.