The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a persistent memory system where historical data is ingested and used to influence agent behavior.
Ingestion points: The agent is instructed to read content from previous sessions using tools like mem_search, mem_context, and mem_get_observation (SKILL.md).
Boundary markers: The protocol does not specify the use of delimiters or specific warnings to segregate retrieved memory content from current instructions.
Capability inventory: Data retrieved from memory is used to answer user queries, recall past solutions, and guide future work.
Sanitization: No sanitization or validation logic is defined to protect against malicious instructions that might be stored in the persistent memory.
[COMMAND_EXECUTION]: The skill documentation references a CLI command for maintenance and tool setup.
Evidence: The 'Fallback' section in SKILL.md recommends running engram setup codex if tools are unavailable, which is a standard procedure for configuring the specific tool integration described.

engram-memory