adk-agent-builder
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions and examples for executing commands via shell, including package installation (pip install google-adk), local agent execution (adk run), and cloud deployment (adk deploy). These are standard operations for a developer tool.
- [PRIVILEGE_ESCALATION]: Documentation in references/errors.md includes guidance on using gcloud commands to grant IAM roles (e.g., roles/aiplatform.user and roles/iam.serviceAccountTokenCreator) to resolve permission issues during Vertex AI deployment. This is standard configuration guidance for Google Cloud services.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided descriptions and requirements to generate code and scaffolds. 1. Ingestion points: User prompts describing agent requirements. 2. Boundary markers: No specific delimiters are prescribed for generated agent instructions. 3. Capability inventory: The skill utilizes Write, Edit, and Bash tools to create files and execute deployment commands. 4. Sanitization: No explicit sanitization of user-provided content is performed before interpolation into generated prompts.
Audit Metadata