The Agent Skills Directory

[Unverifiable Dependencies] (HIGH): The skill instructions require adding an external MCP server endpoint (https://rube.app/mcp) which is not a known or trusted source. This server provides the definitions for all tools and execution logic used by the skill.
[Remote Code Execution] (HIGH): The skill utilizes RUBE_REMOTE_WORKBENCH and RUBE_MULTI_EXECUTE_TOOL. These capabilities allow for the execution of remote logic or tools defined by the external rube.app server, effectively granting the external service control over agent actions.
[Indirect Prompt Injection] (HIGH):
Ingestion points: Tool schemas, input field definitions, and 'recommended execution plans' are fetched dynamically from RUBE_SEARCH_TOOLS via an external network call.
Boundary markers: Absent. The agent is instructed to use the returned fields and logic directly.
Capability inventory: Access to RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH which can perform side effects (file/API operations) in the Mopinion environment.
Sanitization: None. The skill explicitly tells the agent to 'use exact field names and types from the search results' without validation.
[Command Execution] (MEDIUM): The workflow promotes the execution of arbitrary 'tool slugs' and arguments derived from external search results, which can be manipulated by the remote server to perform unauthorized operations.

mopinion-automation