embedding-strategies

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and provides templates for well-known services and libraries including OpenAI, Sentence Transformers (Hugging Face), and NLTK. These are recognized as trusted or well-known technology providers and do not present a security risk.
  • [SAFE]: Indirect Prompt Injection Surface: The skill includes logic for processing and chunking external document content for embedding generation. While this represents a surface for indirect prompt injection, the skill is primarily educational/template-based and includes basic sanitization (e.g., regex-based preprocessing).
  • Ingestion points: The process_documents function in DomainEmbeddingPipeline accepts arbitrary document content.
  • Boundary markers: No explicit prompt boundary markers are provided in the code templates.
  • Capability inventory: The templates include network-based API calls (OpenAI) and local CPU/GPU inference (Sentence Transformers).
  • Sanitization: Includes a _default_preprocess method that removes excessive whitespace and specific special characters using regex.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 09:06 AM