The Agent Skills Directory

[SAFE]: The skill documentation and Python scripts describe and implement standard data structures for state persistence. No malicious patterns, hidden commands, or unauthorized data access were detected.
[PROMPT_INJECTION]: The memory architecture allows for the ingestion of external data that is later retrieved into the agent's context. This represents an indirect prompt injection surface if the memory system is populated with untrusted content.
Ingestion points: Data is stored via the IntegratedMemorySystem.store_fact function in scripts/memory_store.py and retrieved in MemoryContextIntegrator.build_context in references/implementation.md.
Boundary markers: The implementation does not include specific delimiters or ignore-instructions for the retrieved data, potentially allowing embedded instructions to be processed by the agent.
Capability inventory: The skill is restricted to memory management and does not have access to subprocesses, network tools, or other dangerous capabilities.
Sanitization: No sanitization or escaping of the stored facts is performed before they are added to the prompt context.

memory-systems