langgraph
Pass
Audited by Gen Agent Trust Hub on Nov 8, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- External Package Installation: The skill references standard installations for langgraph and @langchain/langgraph. These are official components of the framework provided by the vendor.
- Untrusted Data Processing Surface: The skill facilitates building agents that process MessagesState, creating a surface for potential indirect prompt injection. As documented in SKILL.md: (1) Ingestion Point: graph.invoke() accepts external messages. (2) Capability: Graph nodes can execute arbitrary logic. (3) Boundary Markers: None explicitly defined in basic examples. (4) Sanitization: Not included in the basic configuration. Developers are encouraged to use the framework's interrupt() feature for human-in-the-loop verification of sensitive operations.
Audit Metadata