jobnimbus-automation
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- EXTERNAL_DOWNLOADS (HIGH): The skill instructs users to configure an untrusted remote MCP server endpoint (
https://rube.app/mcp). This is an unverifiable external dependency that manages the core logic and tool definitions for the skill. - PROMPT_INJECTION (HIGH): The skill is highly vulnerable to Indirect Prompt Injection (Category 8). It ingests data from Jobnimbus (untrusted external content) and possesses powerful 'write' capabilities via
RUBE_MULTI_EXECUTE_TOOL. There are no documented boundary markers, delimiters, or sanitization procedures to prevent malicious CRM data from hijacking the agent's behavior. - REMOTE_CODE_EXECUTION (HIGH): The use of
RUBE_REMOTE_WORKBENCHand dynamic tool execution orchestrated by an untrusted remote endpoint (rube.app) constitutes a remote code execution risk, as the server can provide and execute arbitrary tool logic within the agent's context. - COMMAND_EXECUTION (HIGH): The tool provides primitives for broad execution (
RUBE_MULTI_EXECUTE_TOOL) based on dynamically discovered schemas from a remote source, which can be exploited if the source or the data it processes is compromised.
Recommendations
- AI detected serious security threats
Audit Metadata