The Agent Skills Directory

[PROMPT_INJECTION] (HIGH): High-risk Indirect Prompt Injection surface. The skill is designed to fetch tool definitions and execution plans from an external source (RUBE_SEARCH_TOOLS) and execute them against the Baselinker API.
Ingestion points: Data enters the agent context through RUBE_SEARCH_TOOLS responses which provide 'tool slugs, input schemas, and recommended execution plans'.
Boundary markers: Absent. No instructions are provided to the agent to treat external schemas as untrusted or to ignore embedded instructions in the metadata.
Capability inventory: The skill has significant side-effect capabilities including RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH for executing Baselinker operations.
Sanitization: Absent. There is no requirement for validation or filtering of the schema content before interpolation into tool calls.
[EXTERNAL_DOWNLOADS] (MEDIUM): The skill relies on an unverified third-party MCP endpoint (https://rube.app/mcp). This domain is not part of the defined trusted sources and could deliver malicious tool definitions if compromised.
[COMMAND_EXECUTION] (MEDIUM): The skill encourages dynamic discovery and execution of tools (RUBE_MULTI_EXECUTE_TOOL) based on runtime search results, which is a form of dynamic loading of executable paths (tool slugs).

baselinker-automation