imgix-automation

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
  • Unverifiable Dependencies & Remote Code Execution (HIGH): The skill requires connecting to an external MCP endpoint at https://rube.app/mcp. This domain is not on the trusted list, meaning the agent is delegating tool definitions and execution logic to an unverified third party.- Indirect Prompt Injection (HIGH): The skill facilitates a high-risk indirect prompt injection surface.
  • Ingestion points: Data enters the agent context through tools interacting with the Imgix API (e.g., image metadata, asset descriptions).
  • Boundary markers: The skill documentation lacks instructions for implementing delimiters or security boundaries to separate untrusted Imgix content from agent instructions.
  • Capability inventory: The skill uses RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH, providing the agent with write and execution capabilities that could be abused if malicious instructions are found in Imgix data.
  • Sanitization: There is no evidence of sanitization or strict schema validation to prevent external content from influencing the agent's behavior.- Dynamic Execution (MEDIUM): The use of RUBE_REMOTE_WORKBENCH with run_composio_tool() implies the ability to dynamically load and execute tool logic at runtime from a remote source.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 01:11 PM