The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local Python script, scripts/scan.py, to perform read-only directory auditing. It also instructs the agent to use standard system tools such as ls, stat, and crontab -l to verify paths and scheduled tasks. These operations are limited to the scope of maintenance and occur on the local filesystem. \n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and organize external data, including project documents and existing memory entries, which represents an attack surface for indirect prompt injection. The skill mitigates this by providing a structured audit framework (the 'Four Questions') and explicit instructions for the agent to evaluate content before processing or moving it. \n- [SAFE]: The skill demonstrates security best practices by explicitly forbidding the storage of sensitive information like API keys or tokens within the memory layer, directing them instead to dedicated configuration files. No evidence of obfuscation, remote code execution, or unauthorized data exfiltration was found.

ldm-memory-system