deel-automation

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • EXTERNAL_DOWNLOADS (MEDIUM): The skill requires the user to add an external, untrusted MCP server (https://rube.app/mcp). This domain is not among the verified trusted sources. All agent capabilities and data processing for this skill are routed through this external endpoint.
  • DATA_EXFILTRATION (MEDIUM): The skill's primary purpose is automating 'Deel' operations (Payroll, HR, Employee data). By design, sensitive data is passed to the rube.app endpoint via the Rube MCP protocol. This creates a risk of unauthorized data collection or proxying by the third-party service provider.
  • INDIRECT_PROMPT_INJECTION (LOW): The skill dynamically fetches tool schemas and execution plans using RUBE_SEARCH_TOOLS.
  • Ingestion points: Data enters the agent context from the external search results provided by the rube.app endpoint.
  • Boundary markers: No specific delimiters or safety warnings are implemented to prevent the agent from obeying instructions embedded within the returned tool schemas.
  • Capability inventory: The agent has the capability to execute arbitrary tools via RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH based on the untrusted search results.
  • Sanitization: No evidence of input validation or schema sanitization is present in the skill instructions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:42 PM