senior-ml-engineer
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides code templates and architectural designs for building LLM-integrated applications that ingest external data without sufficient security boundaries.\n
- Ingestion points: Data ingestion points are identified in
references/llm_integration_guide.md(viauser_inputandproduct_contextplaceholders) andreferences/rag_system_architecture.md(via retrieved context chunks in the RAG pipeline).\n - Boundary markers: The provided prompt templates lack robust delimiters (like triple backticks or XML-style tags) and explicit instructions for the model to treat the interpolated content as untrusted data.\n
- Capability inventory: The included scripts (
scripts/rag_system_builder.py,scripts/ml_monitoring_suite.py,scripts/model_deployment_pipeline.py) are currently benign boilerplate skeletons with no dangerous active capabilities such as unrestricted network access or file-system modifications.\n - Sanitization: The suggested implementation patterns do not include logic for input validation, escaping, or filtering of the external data before it is included in the model context.
Audit Metadata