The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill includes instructions for deploying sign-in enforcement that require administrative or root access to the target host.\n
Linux: The deployment guide in references/sign-in-enforcement-deploy.md uses sudo to write configuration to /etc/docker-sbx/config.json.\n
Windows: The deployment guide in references/sign-in-enforcement-deploy.md uses registry paths in HKLM (HKEY_LOCAL_MACHINE), which requires elevated privileges.\n
macOS: Mentions using MDM-deployed configuration profiles for policy enforcement, which operate at a system level.\n- [COMMAND_EXECUTION]: The skill provides various administrative commands for local policy management and system configuration.\n
Commands include sbx policy reset and sbx reset, which terminate running sandboxes and modify daemon states.\n
The skill documents usage of defaults write (macOS) and New-ItemProperty (Windows) to enforce security settings.\n- [INDIRECT_PROMPT_INJECTION]: The skill manages sandbox governance through external policies, creating an attack surface for indirect prompt injection.\n
Ingestion points: Policies fetched from the Governance API or Docker Home (SKILL.md).\n
Boundary markers: Absent.\n
Capability inventory: Filesystem mount access, network egress control, and audit logging configuration.\n
Sanitization: Absent.

governing-sbx-fleets