how-we-work
Pass
Audited by Gen Agent Trust Hub on Nov 11, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes a specialized Python script (
hww.py) to manage markdown synchronization within projects. It also provides instructions for standard development tools such asgit,ssh, anduv. These tools are used appropriately for project management and experiment tracking. - [EXTERNAL_DOWNLOADS]: The documentation includes a reference to the official installation script for the
uvtool fromastral.sh. Astral is a well-known and reputable technology company, and the download is part of standard environment setup instructions. - [DATA_EXFILTRATION]: The skill includes a notification mechanism using the
Barkpush notification service. This is an optional, user-configured feature for monitoring long-running unattended tasks. Configuration (such as the server URL and device key) is stored in a local user directory, and the transmitted data is limited to task status updates. - [PRIVILEGE_ESCALATION]: The instructions for using remote training servers explicitly forbid the use of
sudoor system-wide package installations by the agent, directing the agent to provide commands for the human user to execute instead. This follows the principle of least privilege. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied markdown files to extract rules for promotion. The instructions require the agent to judge each entry manually against established criteria, maintaining a human-in-the-loop safety model for documentation changes.
Audit Metadata