The Agent Skills Directory

[Dynamic Execution] (MEDIUM): The skill workflow relies on RUBE_SEARCH_TOOLS to fetch tool slugs and schemas from https://rube.app/mcp at runtime. The agent then executes these tools via RUBE_MULTI_EXECUTE_TOOL. Since the logic and parameters are defined by an external, non-trusted source and executed dynamically, it poses a medium risk of remote logic control.
[Indirect Prompt Injection] (LOW): The skill identifies a surface for tool output poisoning. 1. Ingestion points: Tool schemas and recommended plans returned by RUBE_SEARCH_TOOLS. 2. Boundary markers: Absent; instructions suggest using exact field names and following recommended plans from the search results. 3. Capability inventory: RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH allow for complex interactions and command-like tool execution. 4. Sanitization: Absent; there is no validation of the schemas or instructions returned by the remote server.
[External Reference] (LOW): The skill requires connection to https://rube.app/mcp. This domain is not included in the list of trusted external sources.

goody-automation