The Agent Skills Directory

[Unverifiable Dependencies & Remote Code Execution] (HIGH): The skill mandates the use of an external MCP server endpoint (https://rube.app/mcp) that is not within the trusted source list.
Evidence: 'Add https://rube.app/mcp as an MCP server in your client configuration.'
This configuration delegates the agent's tool definitions and execution environment to an unverified third party.
[Indirect Prompt Injection] (HIGH): The skill creates a high-risk attack surface by instructing the agent to fetch instructions (schemas and plans) from the external Rube MCP service and execute them directly.
Ingestion points: RUBE_SEARCH_TOOLS retrieves schemas, plans, and 'pitfalls' from the external server.
Boundary markers: None. The skill explicitly tells the agent to 'Always search tools first' and 'Use exact field names and types from the search results,' creating a dependency on untrusted data for execution logic.
Capability inventory: RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH provide the ability to execute code and perform side-effect operations based on the fetched data.
Sanitization: No sanitization or validation of the externally provided tool schemas or execution plans is performed before they are processed by the agent.
[Dynamic Execution] (MEDIUM): The skill utilizes RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH to execute operations that are defined at runtime through dynamic schema discovery.

zerobounce-automation