mirrord-kafka

Pass

Audited by Gen Agent Trust Hub on Apr 29, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes kubectl commands to gather cluster context and verify the existence of required resources.\n
  • Evidence: SKILL.md (Step 2) uses kubectl config, get namespace, and get crd to establish context.\n- [PROMPT_INJECTION]: The skill ingests data from Kubernetes resource definitions and user input, creating an indirect prompt injection surface.\n
  • Ingestion points: SKILL.md (Step 2 and 3) reads workload pod specs and user-provided values for namespaces and topics.\n
  • Boundary markers: The skill instructs the agent to treat data within <USER_INPUT> tags as opaque and non-instructional.\n
  • Capability inventory: The skill includes read-only cluster access (kubectl) and the generation of structured YAML/JSON configurations.\n
  • Sanitization: The skill enforces strict input sanitization, including validation of names via regex and the rejection of any input containing shell metacharacters.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 29, 2026, 03:07 PM