llm-app-patterns

Pass

Audited by Gen Agent Trust Hub on Apr 7, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is entirely educational, providing best-practice code snippets for LLM orchestration and reliability.
  • [SAFE]: Explicitly includes a security section on Guardrails, demonstrating how to implement input validation and detect common prompt injection patterns. The strings flagged by static detectors are used as detection patterns within this security code, not as malicious instructions.
  • [SAFE]: No hardcoded secrets, dangerous network operations, or unexpected command executions were found.
  • [SAFE]: Dependencies mentioned in the code (Anthropic SDK, Express, AJV, eventsource-parser) are well-known, reputable packages used for their intended purposes.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 7, 2026, 10:29 AM
Security Audit — agent-trust-hub — llm-app-patterns