ai-rag-pipeline

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill is designed to orchestrate AI tasks using the infsh command-line interface, which is the intended behavior for the vendor's platform.
  • [EXTERNAL_DOWNLOADS]: The documentation references installation commands and assets from official repositories and domains owned by the vendor (inference.sh).
  • [PROMPT_INJECTION]: The Retrieval-Augmented Generation (RAG) pattern demonstrated in the skill presents a surface for indirect prompt injection by processing external web content. (1) Ingestion points: Web data is retrieved via tools like tavily/search-assistant and exa/search and stored in shell variables. (2) Boundary markers: The prompt templates do not use specific delimiters to isolate external content from instructions. (3) Capability inventory: The skill utilizes infsh app run to execute LLM models. (4) Sanitization: Retrieved content is interpolated directly into prompts without pre-filtering, which is a standard architectural characteristic of basic RAG implementations.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 09:31 PM