langchain-tool-calling

Pass

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill consists of educational documentation and code snippets for the LangChain framework. No malicious patterns, obfuscation, or unauthorized data access were identified.
  • [INDIRECT_PROMPT_INJECTION]: The skill documents the standard agentic pattern of feeding external tool outputs back into the language model's context. This creates a potential surface for indirect prompt injection if the external data source is untrusted.
  • Ingestion points: The documentation demonstrates appending tool execution results directly to the conversation history (messages.push(...toolResults)) in the manual execution and conversation examples.
  • Boundary markers: The code examples do not include explicit delimiters (like XML tags or markdown blocks) or system instructions to ignore instructions embedded within tool results.
  • Capability inventory: The demonstrated tools include weather queries, news queries, and search functions, which involve processing external data.
  • Sanitization: No input validation or output sanitization is demonstrated for the tool results before they are processed by the model.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 16, 2026, 12:30 AM
Security Audit — agent-trust-hub — langchain-tool-calling