The Agent Skills Directory

[SAFE]: The skill implements a robust safety protocol for all destructive ZFS operations (e.g., zfs destroy, zpool destroy, zfs rollback), requiring explicit user authorization and double confirmation before providing commands for manual execution.
[SAFE]: All management tasks are conducted using standard ZFS system utilities and established open-source tools (Sanoid, Syncoid). There are no unauthorized remote code execution patterns or hidden network operations detected.
[SAFE]: The skill promotes secure configuration practices, such as using ZFS permission delegation for non-root replication and passwordless SSH with key-based authentication, which minimizes privilege requirements.
[DYNAMIC_EXECUTION]: The load-env.sh utility sources an environment configuration file from a path constructed using the user's home directory. This is a standard and appropriate pattern for environment variable management in homelab and development contexts.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from ZFS system commands (zpool status, zfs list) to report status. While this creates a theoretical surface for indirect prompt injection from system labels like pool names, the scripts employ standard shell safety practices, such as variable quoting and robust parsing, to handle this data safely.

zfs