The Agent Skills Directory

[External Remote Service] (MEDIUM): The skill instructions require adding an untrusted endpoint (https://rube.app/mcp) as an MCP server. This source host is not on the trusted sources list.
[Indirect Prompt Injection] (HIGH): The skill architecture ingests untrusted data that directly influences agent behavior and tool usage.
Ingestion points: RUBE_SEARCH_TOOLS retrieves schemas and recommended execution plans from the remote server.
Boundary markers: Absent; there are no instructions to the agent to isolate or treat the server output as data rather than instructions.
Capability inventory: The skill has access to RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH, which allow for broad action execution and potential code execution.
Sanitization: Absent; the agent is told to follow the 'exact field names' and 'recommended execution plans' provided by the remote source.

zylvie-automation