ark-sdk-development
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [Indirect Prompt Injection] (MEDIUM): The skill describes a type generation pipeline that is vulnerable to indirect injection via malicious Kubernetes CRD definitions.\n
- Ingestion points: CRD YAML files in
ark/config/crd/bases/*.yaml.\n - Boundary markers: None specified in the instructions.\n
- Capability inventory: Execution of
makeandnpmbuild scripts, Python script execution (crd_to_openapi.py), and file system modifications (overlay copying).\n - Sanitization: No mention of validation or sanitization for the source CRD files before they are processed by generators.\n- [Command Execution] (LOW): The instructions require the agent to execute various shell commands including
make,npm, andcp. While standard for local development, these represent significant capabilities that could be abused if the underlying build scripts or input data are compromised.
Audit Metadata