The Agent Skills Directory

[SAFE]: The skill utilizes platform-native tools (episodic_search, episodic_read_window, episodic_read_session) for their intended purpose of retrieving historical conversation context. No evidence of malicious behavior, data exfiltration, or unauthorized access was found.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow for the agent to ingest content from past conversations. While this creates a surface for indirect prompt injection (where a previous conversation might contain instructions aimed at a future session), this is a characteristic of memory-augmented agents and not a flaw in the skill itself.
Ingestion points: Past conversation history is retrieved through episodic_search, episodic_read_window, and episodic_read_session as described in SKILL.md.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions for the retrieved data, relying instead on the agent's ability to distinguish between historical records and current instructions.
Capability inventory: The agent has the capability to read and integrate historical session data into its active context.
Sanitization: The skill relies on a platform-level opt-out mechanism (DO NOT INDEX THIS CHAT) rather than runtime content sanitization.

remembering-conversations