The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from recent conversations and daily notes to update long-term memory.
Ingestion points: Data is pulled from the file system via read_note, search_notes, and recent_activity calls targeting daily logs and memory files.
Boundary markers: The skill does not define clear delimiters or specific instructions for the agent to disregard instructions embedded within the raw notes during reflection.
Capability inventory: The skill is restricted to reading local files and writing consolidated summaries to MEMORY.md and daily notes; it does not exhibit network exfiltration or arbitrary command execution capabilities.
Sanitization: The instructions do not specify any validation, filtering, or sanitization of the content before it is processed and written back to memory files.

memory-reflect