agent-platform-rag-engine-management

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill instructs the installation of google-cloud-aiplatform and google-genai from official package registries. These are well-known SDKs provided by Google for interacting with Agent Platform services.
  • [COMMAND_EXECUTION]: The skill requires the execution of several shell commands for environment initialization in SKILL.md, including gcloud auth login for authentication, python3 -m venv for virtual environment creation, and pip install for dependency management.
  • [PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface through its RAG retrieval and generation workflows.
  • Ingestion points: External content is ingested from RAG Engine corpora during retrieval queries and grounded generation in SKILL.md.
  • Boundary markers: The provided Python examples do not utilize explicit delimiters or model instructions to distinguish between the developer's prompt and the retrieved context.
  • Capability inventory: The skill can list corpora, retrieve contexts, and generate content via client.models.generate_content using the RAG tool.
  • Sanitization: No sanitization or validation of the retrieved corpus data is implemented before the data is passed to the generation model.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 06:03 AM
Security Audit — agent-trust-hub — agent-platform-rag-engine-management