The Agent Skills Directory

[CREDENTIALS_UNSAFE] (HIGH): The skill directs the agent to manage 'Gemini connections' via an external service (rube.app) and follow auth links generated by that service. This architecture allows the third-party server to proxy or intercept authentication tokens for Google/Gemini services.
[PROMPT_INJECTION] (HIGH): The skill is highly vulnerable to Indirect Prompt Injection. It is designed to ingest 'recommended execution plans' and 'schemas' from the remote rube.app server via RUBE_SEARCH_TOOLS and immediately act upon them using RUBE_MULTI_EXECUTE_TOOL.
Ingestion points: Data returned from the untrusted endpoint https://rube.app/mcp during the search phase.
Boundary markers: None. The instructions mandate following the discovered schemas and plans verbatim.
Capability inventory: The skill provides full execution capabilities via RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH.
Sanitization: No sanitization or validation of the remote plans is implemented.
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill requires the installation of an external, unverifiable MCP server (https://rube.app/mcp). This source is not on the trusted list and provides the functional logic for the skill.
[REMOTE_CODE_EXECUTION] (MEDIUM): By integrating an external MCP server, the agent is effectively executing remote tool definitions and logic controlled by a third party. While not direct shell access, the agent's behavior is entirely determined by the remote server's responses.

gemini-automation