The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use subprocess.run to interact with the local miniqmt-cli tool. This is the primary intended operation for wrapping the CLI functionality.
[COMMAND_EXECUTION]: Utilizes standard system commands such as ssh for port forwarding, schtasks for Windows daemon management, and launchctl for macOS tunnel persistence. These operations are consistent with the described client-server architecture and are used for legitimate system administration tasks.
[CREDENTIALS_UNSAFE]: No hardcoded credentials were found. The skill follows best practices by instructing users to manage account configurations in local .toml files and explicitly mentions that account IDs are masked in output.
[PROMPT_INJECTION]: The instructions are purely technical and operational. No attempts to override AI behavior, bypass safety filters, or extract system prompts were detected.
[DATA_EXFILTRATION]: All network operations (FastAPI daemon) are directed to 127.0.0.1:8765 over an SSH tunnel controlled by the user. There are no signs of data being sent to unauthorized external domains.
[EXTERNAL_DOWNLOADS]: While the skill mentions setup scripts (bootstrap.ps1, deploy.sh), it does not provide remote URLs or commands to download code from untrusted external sources.

miniqmt-cli