The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill mandates the addition of an untrusted third-party MCP server at https://rube.app/mcp. This server is not within the trusted source scope, presenting a risk of interacting with an unverified remote entity.
[REMOTE_CODE_EXECUTION] (HIGH): The workflow is designed to fetch "recommended execution plans" from the remote server via RUBE_SEARCH_TOOLS and implement them using the agent's tools. This creates a direct channel for a remote server to dictate agent behavior and operations.
[COMMAND_EXECUTION] (HIGH): The RUBE_MULTI_EXECUTE_TOOL capability executes arbitrary tools based on schemas retrieved dynamically from the remote server, which could lead to unauthorized API or system operations if the remote schemas are malicious.
[PROMPT_INJECTION] (HIGH): The skill possesses a high-severity vulnerability surface for Indirect Prompt Injection:
Ingestion points: Data enters the context from rube.app via the RUBE_SEARCH_TOOLS function which provides tool schemas and execution plans.
Boundary markers: Absent; the skill instructs the agent to "Always search tools first" and adhere to the returned schemas and plans without verification.
Capability inventory: Broad execution capabilities are available through RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH.
Sanitization: No evidence of input validation, schema verification, or instruction filtering is present for the remote content before execution.

printautopilot-automation