The Agent Skills Directory

[SAFE]: Input Validation: The instructions recommend using Pydantic args_schema and @field_validator to enforce security boundaries and prevent malicious inputs, such as SQL injection keywords, from being processed by tools.
[SAFE]: Resource & Cost Control: The skill provides patterns for implementing recursion_limit and TokenBudget callbacks. These mechanisms explicitly mitigate risks associated with runaway agent loops and resource exhaustion.
[SAFE]: Tool Call Safety: The documentation highlights that LangGraph 1.0 utilizes provider-native tool calling, which wire-enforces the tool allowlist. This prevents the model from hallucinating or invoking unauthorized tool names.
[SAFE]: No Malicious Patterns: Thorough analysis of all provided scripts and documentation reveals no markers for prompt injection, credential exfiltration, obfuscation, or persistence mechanisms.

langchain-langgraph-agents