The Agent Skills Directory

[SAFE]: The skill provides educational content and Python reference code for building tiered memory systems. The implementation uses standard libraries (numpy, json, hashlib) and does not perform network operations or access sensitive system files.\n- [PROMPT_INJECTION]: The MemoryContextIntegrator class in references/implementation.md provides a functional pattern for indirect prompt injection by retrieving stored memories and injecting them into the context window.\n
Ingestion points: The build_context method in references/implementation.md ingests user-provided task strings to trigger entity-based memory retrieval.\n
Boundary markers: Absent. The logic uses simple newline delimiters (\\n\\n) to separate retrieved memories from the primary context, which may allow retrieved instructions to influence agent behavior.\n
Capability inventory: No dangerous capabilities (shell execution, network calls, or file writes) are present in the provided scripts.\n
Sanitization: No sanitization or validation of the retrieved memory content is performed before interpolation into the prompt.

memory-systems