The Agent Skills Directory

[PERSISTENCE]: The skill automates the creation of system persistence mechanisms, specifically launchd agents on macOS and systemd user units on Linux, to ensure the server remains active as a background process.
[PRIVILEGE_ESCALATION]: Host-level networking configuration relies on the execution of commands with elevated privileges, such as using sudo to configure Tailscale operator settings and network initialization.
[COMMAND_EXECUTION]: The skill utilizes the exec tool to perform extensive administrative operations, including service deployment, network tunneling configuration, and the generation of authentication credentials.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface where instructions or data from remote cloud agents and external devices are ingested into the local agent's context.
Ingestion points: Incoming MCP requests from remote clients connected via the Tailscale or Funnel endpoints defined in SKILL.md.
Boundary markers: Enforcement of least-privilege access using predefined profiles such as memory-reader and memory-writer to limit the scope of remote operations.
Capability inventory: The agent has access to the exec tool, which is used for managing the host's server and networking state.
Sanitization: Input validation and command escaping are delegated to the gbrain server's administrative and client APIs.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download and installation of the Tailscale binary from its official domain (tailscale.com) to provide the encrypted transport layer for remote connectivity.

remote-mcp