The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to handle user-supplied messages through LangGraph's MessagesState. The provided templates include architectural patterns (like supervisors and triage agents) that manage task routing. While this creates an ingestion surface for untrusted data, the skill includes best practices like recursion guards (handoff_count) and structured output schemas to maintain control flow integrity.
[DYNAMIC_EXECUTION]: The scripts/ directory contains Python scripts such as lg-supervisor-scaffold.py and lg-swarm-scaffold.py that generate boilerplate project structures. These scripts use static string templates to produce code files for developer use and do not execute dynamically generated code at runtime or incorporate external untrusted input into execution. The use of .compile() in the templates refers to the LangGraph library's method for finalizing graph structures, which is standard framework behavior.
[EXTERNAL_DOWNLOADS]: The skill documentation and requirements point to well-known, official packages including langgraph, langchain, and langsmith from established registries. All referenced external resources are from trusted organizations or well-known technology services.

langgraph