coassemble-automation

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • EXTERNAL_DOWNLOADS (MEDIUM): The skill directs the user to connect to an external MCP endpoint at https://rube.app/mcp. This domain is not a trusted source per the analysis guidelines, meaning the tool definitions and code it provides are unverifiable.\n- COMMAND_EXECUTION (MEDIUM): The skill utilizes RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH to perform tasks. These tools execute operations determined by schemas and instructions fetched at runtime from the untrusted remote service, creating a surface for dynamic execution of unknown logic.\n- PROMPT_INJECTION (LOW): The skill is susceptible to indirect prompt injection due to its handling of external data from tool outputs.\n
  • Ingestion points: Data enters the context via RUBE_SEARCH_TOOLS results and the Coassemble toolkit outputs.\n
  • Boundary markers: Absent. The instructions do not provide delimiters or warnings to ignore embedded instructions in the tool responses.\n
  • Capability inventory: Significant capabilities are available through the RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH functions.\n
  • Sanitization: Absent. The agent is instructed to use schemas and field names exactly as returned from the remote search results without validation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:36 PM