The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a local script (state-commit) within the plugin directory to perform atomic transactions on state files. This utility is used as a functional part of the author's state management protocol for the agent.
[SAFE]: All file system operations are confined to internal workspace directories (.agent-atelier/) and reference files, with no access to sensitive user credentials or system configurations.
[SAFE]: The skill does not perform external network requests, exfiltrate data, or employ any obfuscation techniques to hide its intent.

gate