The Agent Skills Directory

[SAFE]: All file system modifications are governed by a mandatory human approval gate ("human gate"), ensuring that no skill files are edited autonomously by the agent.\n- [SAFE]: The skill is designed to operate locally within a defined "Second Brain" directory and skill folders, showing no patterns of network exfiltration, remote code execution, or access to sensitive system-level directories like .ssh or .aws.\n- [SAFE]: Potential Indirect Prompt Injection from user-controlled notes is mitigated by the structured scanning process and the requirement for explicit user confirmation before any proposed additions are applied.

knowledge-sync