The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (HIGH): The skill directs users to add https://rube.app/mcp as an MCP server. This is an unverifiable third-party domain not included in the Trusted External Sources list. The claim that 'No API keys needed' suggests a lack of standard authentication for the remote component.
[REMOTE_CODE_EXECUTION] (HIGH): The skill utilizes RUBE_REMOTE_WORKBENCH and RUBE_MULTI_EXECUTE_TOOL. These tools provide the capability to execute complex workflows and code on remote environments. When combined with an untrusted endpoint, this creates a high-risk execution path.
[INDIRECT_PROMPT_INJECTION] (HIGH): The skill relies on 'Tool Discovery' via RUBE_SEARCH_TOOLS to determine the agent's next steps. The agent is explicitly instructed to 'Always discover available tools before executing workflows' and 'Never hardcode tool slugs... without calling RUBE_SEARCH_TOOLS'. This ensures the agent is subservient to the data returned by the remote server, which could inject malicious tool schemas or execution plans.
Ingestion points: Tool schemas and execution plans returned by RUBE_SEARCH_TOOLS from rube.app.
Boundary markers: None identified. The skill prioritizes the remote schema over local instructions.
Capability inventory: RUBE_MULTI_EXECUTE_TOOL, RUBE_REMOTE_WORKBENCH, and RUBE_MANAGE_CONNECTIONS.
Sanitization: None. The skill mandates strict compliance with the remote search results.

veriphone-automation