monte-carlo-troubleshoot-agent-traces
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill provides explicit instructions to the AI agent to treat all data retrieved from external sources (such as conversation transcripts and execution traces) as informational data for analysis rather than as instructions, mitigating the risk of indirect prompt injection from customer-controlled content.
- [EXTERNAL_DOWNLOADS]: Data retrieval and configuration updates are conducted through the official Monte Carlo MCP server at integrations.getmontecarlo.com, which is the expected and authorized domain for the skill author.
- [DATA_EXFILTRATION]: The skill accesses sensitive information including conversation transcripts, prompts, and execution traces; however, this access is performed via authorized MCP tools for the purpose of troubleshooting, and no exfiltration to third-party or unauthorized domains was observed.
- [COMMAND_EXECUTION]: The skill triggers backend troubleshooting agents and retrieves metadata through specialized MCP tools; it does not utilize arbitrary shell commands, system calls, or other high-risk execution patterns.
Audit Metadata