llm-app-patterns

Pass

Audited by Gen Agent Trust Hub on Jul 28, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill documents patterns for Retrieval-Augmented Generation (RAG) and agents that ingest external data into prompts. It provides templates such as RAG_PROMPT_TEMPLATE and REACT_PROMPT which interpolate context and tool observations. These patterns represent a standard attack surface for indirect prompt injection, and the skill includes basic instructional guardrails to mitigate this risk.
  • Ingestion points: The skill defines prompt templates in SKILL.md that accept external data via {context} and {observation} placeholders.
  • Boundary markers: The provided RAG template includes specific instructions to the model to "Answer the user's question based ONLY on the following context."
  • Capability inventory: The agent patterns describe the use of search and calculation tools.
  • Sanitization: As an architectural guide, the snippets demonstrate structure rather than specific input validation or sanitization logic.
  • [EXTERNAL_DOWNLOADS]: The skill includes references to industry-standard resources and documentation, including the Anthropic Cookbook, LangChain, LlamaIndex, and the Dify platform.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 28, 2026, 05:45 AM
Security Audit — agent-trust-hub — llm-app-patterns