rag-architect

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill integrates with multiple well-known technology services and libraries, including OpenAI, Cohere, Qdrant, Pinecone, Weaviate, and Hugging Face's Sentence Transformers, for embedding generation and vector storage.
  • [PROMPT_INJECTION]: The skill describes an architecture for processing external documents into an LLM context, which presents an inherent surface for Indirect Prompt Injection.
  • Ingestion points: Document ingestion occurs during chunking and hybrid search workflows defined in SKILL.md and references/chunking-strategies.md.
  • Boundary markers: The provided implementation examples do not explicitly demonstrate the use of XML tags, delimiters, or system instructions to ignore embedded commands within the retrieved text segments.
  • Capability inventory: The skill includes capabilities for network API calls (OpenAI, Cohere) and vector database operations (upsert, search) which could be influenced by malicious content in ingested documents.
  • Sanitization: There are no documented steps for sanitizing or filtering retrieved content before it is interpolated into the final prompt for the generator model.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 09:32 AM
Security Audit — agent-trust-hub — rag-architect