The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes a preflight check that provides a command to install the Cube CLI: curl -fsSL https://raw.githubusercontent.com/cube-js/cube/master/install-cli.sh | sh. This script is fetched from the official GitHub repository belonging to the skill's vendor.
[COMMAND_EXECUTION]: The skill relies on the execution of the cube command-line tool to perform all its functions, including listing workspace content, retrieving report definitions, and inspecting notification recipients.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external Cube deployments, which represents a potential attack surface.
Ingestion points: The agent retrieves untrusted data from a Cube workspace via commands like cube reports get <report> and cube workspace list <deployment> as described in SKILL.md.
Boundary markers: There are no explicit boundary markers or instructions to the agent to disregard instructions embedded within the retrieved report data or queries.
Capability inventory: The agent possesses the capability to execute shell commands and interact with the filesystem via the cube CLI across all described workflows.
Sanitization: The skill does not provide methods for sanitizing or validating the content of the saved queries or reports retrieved from the external source before the agent processes them.

cube-explore-content