The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill documentation includes instructions to download and execute an installation script from the vendor's domain to set up host machines.
Evidence: curl -fsSL https://cua.ai/install.sh | sh -s -- -y --select spaces,host --no-onboarding in the host setup section.
[COMMAND_EXECUTION]: The skill enables the agent to run arbitrary shell commands within managed 'Spaces'.
Evidence: space_bash tool for running commands with optional timeouts.
[PERSISTENCE]: The instructions for setting up a host machine involve the use of a macOS LaunchAgent to ensure the cua-spacesd daemon persists across user sessions.
Evidence: "cua-spacesd runs as a LaunchAgent in the Aqua session."
[INDIRECT_PROMPT_INJECTION]: The skill interacts with external environments by reading command outputs and files, creating a potential surface for instructions embedded in external data to influence the agent.
Ingestion points: Output from space_bash, content of files retrieved via download, and results from call_tool within the Space.
Boundary markers: Not explicitly defined in the tool definitions.
Capability inventory: Shell execution (space_bash), file creation (space_write), file transfer (upload/download), and network sharing (hotspot_start).
Sanitization: Instructions include a requirement to get explicit user acknowledgement before moving sensitive app sessions via teleport_app.

cua-spaces