The Agent Skills Directory

[SAFE]: The skill establishes a protocol for using memory and personalization tools to improve agent responses.
It utilizes platform-integrated tools save_memory and recall_memories for state management.
The instructions provide proactive privacy guidelines, explicitly telling the agent to exclude health, financial, and identity details from long-term storage unless specifically directed by the user.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and storing user-provided data, which creates a potential surface for indirect prompt injection.
Ingestion points: User requests and choices are captured and stored as "facts" via the save_memory tool.
Boundary markers: The skill does not define specific delimiters for stored memory content, although it prioritizes current session instructions over saved defaults.
Capability inventory: The skill utilizes save_memory for persistence and recall_memories for context retrieval.
Sanitization: Includes negative constraints to prevent the storage of sensitive data categories (health, financial, identity).

memory-personalization