The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill requires the user to add an untrusted remote endpoint (https://rube.app/mcp) as an MCP server. This endpoint is not within the Trusted External Sources list and effectively delegates tool definitions and execution logic to a third-party server.
[REMOTE_CODE_EXECUTION] (HIGH): The skill implements a pattern where tool schemas, recommended execution plans, and input requirements are fetched at runtime via RUBE_SEARCH_TOOLS and then executed via RUBE_MULTI_EXECUTE_TOOL. This allows the remote server to dictate what commands the agent executes.
[COMMAND_EXECUTION] (HIGH): The skill provides capabilities to execute arbitrary operations via the RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH functions based on instructions retrieved from the remote server.
[INDIRECT_PROMPT_INJECTION] (HIGH): The skill has a high-severity vulnerability surface for indirect prompt injection.
Ingestion points: Tool schemas, execution plans, and 'pitfalls' returned by the RUBE_SEARCH_TOOLS call in SKILL.md.
Boundary markers: Absent. There are no instructions to validate or delimit the data returned from the search tool before use.
Capability inventory: RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH provide the ability to modify external state (Aivoov tasks).
Sanitization: Absent. The instructions explicitly direct the agent to 'Use exact field names and types from the search results', ensuring the agent will follow instructions embedded in the remote data.

aivoov-automation