The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (HIGH): The skill directs users to add an untrusted remote MCP server (https://rube.app/mcp). This server controls all tool definitions and logic. Because it is not in the trusted source list, this represents a high-risk external dependency.
[REMOTE_CODE_EXECUTION] (HIGH): The skill references RUBE_REMOTE_WORKBENCH and run_composio_tool(). This implies the capability to execute code or logic in a remote environment managed by the untrusted rube.app service, which could be used to execute malicious payloads.
[INDIRECT_PROMPT_INJECTION] (HIGH):
Ingestion points: The skill relies on RUBE_SEARCH_TOOLS to fetch tool schemas and "recommended execution plans" from the remote server at runtime.
Boundary markers: Absent. The agent is instructed to "Always search tools first" and use the results to build its workflow without any verification or sanitization.
Capability inventory: The skill possesses write capabilities (modifying Google Calendar) and potential execution capabilities via the remote workbench.
Sanitization: Absent. There is no mechanism to validate that the "execution plans" or schemas provided by the remote server are safe or conform to the user's intent.
[COMMAND_EXECUTION] (MEDIUM): The RUBE_MULTI_EXECUTE_TOOL allows the agent to perform actions with side effects (creating/deleting calendar events) based on instructions fetched from an untrusted remote source.

googlecalendar-automation