The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands like grep, sed, and pwd to manage paths and search for existing project memory entries. These are standard operations for maintaining project-specific configuration files within the agent's workspace.
[INDIRECT_PROMPT_INJECTION]: The skill persists user-provided information into a MEMORY.md file that is read by the agent in future sessions, which constitutes an indirect prompt injection surface.
Ingestion points: User input provided via the /si:remember command is written to the filesystem.
Boundary markers: The data is stored in a markdown list format; however, there are no specific boundary markers or instructions to treat this stored content as untrusted during retrieval.
Capability inventory: The skill executes file writes and local shell searches.
Sanitization: No validation or sanitization is performed on the input before it is saved to the persistent memory file.

remember