The Agent Skills Directory

[SAFE]: The skill consists of documentation and code snippets for utilizing LangGraph's memory features. All referenced libraries are official packages from the LangChain ecosystem.
[CREDENTIALS_UNSAFE]: The code example for PostgreSQL storage uses a generic placeholder string ('postgresql://user:pass@localhost/db') intended for local development. It does not contain actual credentials or sensitive environment data.
[INDIRECT_PROMPT_INJECTION]: The skill provides a template for agents to ingest and retrieve persistent user data (memories). While this creates a potential attack surface if retrieved content is not sanitized before inclusion in prompts, it is a standard implementation pattern for memory-enabled agents.
Ingestion points: Data is retrieved via store.get() and store.search() in smartNode and myNode (SKILL.md).
Boundary markers: The provided examples do not demonstrate the use of XML delimiters or instructions to ignore embedded commands within stored memories.
Capability inventory: The code snippets demonstrate file-like persistence via database connectors (PostgresStore) and state management.
Sanitization: No explicit sanitization or validation of the memory content is shown in the educational snippets.

langgraph-memory