The Agent Skills Directory

PROMPT_INJECTION (LOW): The skill contains a surface for Indirect Prompt Injection. It is designed to dynamically ingest tool schemas and execution plans from the RUBE_SEARCH_TOOLS tool at runtime.
Ingestion points: External data enters the context via the output of RUBE_SEARCH_TOOLS (file: SKILL.md).
Boundary markers: Absent. There are no instructions to the agent to treat retrieved tool schemas as untrusted or to ignore embedded instructions within them.
Capability inventory: The skill utilizes RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH, which provide the capability to execute operations and manage connections.
Sanitization: No sanitization or validation of the remote schemas is performed by the skill logic.
EXTERNAL_DOWNLOADS (LOW): The skill directs users to configure an external MCP server endpoint (https://rube.app/mcp). This domain is not listed in the Trusted External Sources. Although this is a configuration step rather than an automated script execution (like curl | bash), it establishes a dependency on an untrusted remote service for the skill's core functionality.

hookdeck-automation