recallai-automation

Warn

Audited by Gen Agent Trust Hub on Feb 19, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • EXTERNAL_DOWNLOADS (MEDIUM): The skill requires the configuration of an external MCP server at https://rube.app/mcp. This domain is not on the list of trusted providers (e.g., Anthropic, OpenAI, Microsoft), representing an unverifiable remote dependency that defines the agent's available tools.
  • COMMAND_EXECUTION (MEDIUM): The skill utilizes RUBE_REMOTE_WORKBENCH, which suggests a remote environment for code or tool execution. Remote workbenches often involve elevated capabilities that could be exploited if the underlying service is compromised.
  • PROMPT_INJECTION (LOW): As a tool designed to process Recallai data (meeting records), the skill is susceptible to indirect prompt injection. Malicious instructions embedded in a meeting transcript or tool discovery metadata could attempt to hijack the agent's logic during the RUBE_MULTI_EXECUTE_TOOL step.
  • Ingestion points: RUBE_SEARCH_TOOLS (tool schemas) and Recallai meeting data processed by various tool slugs.
  • Boundary markers: Absent; no instructions are provided to the agent to treat tool output or search results as untrusted data.
  • Capability inventory: RUBE_MULTI_EXECUTE_TOOL, RUBE_REMOTE_WORKBENCH, and RUBE_MANAGE_CONNECTIONS allow for complex, state-changing operations.
  • Sanitization: No sanitization or validation of dynamic tool arguments is specified before they are passed to the execution tools.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 19, 2026, 04:04 AM