The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill relies on dynamic tool discovery, meaning the agent's execution logic is determined by data retrieved from an external service (RUBE_SEARCH_TOOLS). This design is part of the standard toolkit operation but technically exposes a surface for indirect instructions through tool schemas.
Ingestion points: Content returned from RUBE_SEARCH_TOOLS used to populate tool calls.
Boundary markers: No explicit delimiters are specified in the markdown to isolate the external tool schema from the agent's processing instructions.
Capability inventory: The skill uses RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH to act upon the discovered schemas.
Sanitization: The instructions recommend using schemas exactly as returned without additional validation layers.

openrouter-automation