The Agent Skills Directory

[SAFE]: The skill provides standard architectural patterns and code implementations for AI agent memory systems.
[PROMPT_INJECTION]: The skill establishes a persistence layer that stores untrusted user data for future retrieval into the agent's context window, creating an indirect prompt injection surface.
Ingestion points: Data enters through log_today and remember in agent_memory.py, store in memory-store.ts, and store in chroma_memory.py.
Boundary markers: Absent. The implementation stores and retrieves raw strings without using delimiters or system instructions to distinguish recalled memory from new instructions.
Capability inventory: The skill enables persistent file writing/reading and database operations (SQLite, ChromaDB) to store context across sessions.
Sanitization: Absent. The implementation does not perform sanitization or validation of the content before storage or recall.

agent-memory