mcp-core-best-practices

Pass

Audited by Gen Agent Trust Hub on Aug 4, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes the AlibabaCloud___CallCLI tool to execute Alibaba Cloud CLI commands. It includes explicit security constraints such as prohibiting shell pipes, redirections, variables, and command substitutions to prevent command injection.\n- [REMOTE_CODE_EXECUTION]: The AlibabaCloud___RunScript tool facilitates the execution of Python code on a remote MCP server. The skill includes a robust set of safety guidelines for the agent, explicitly forbidding the use of eval, exec, subprocess, and network operations to ensure the scripts remain within the intended sandbox.\n- [REMOTE_CODE_EXECUTION]: The AlibabaCloud___RunIaC tool allows for the remote execution of Infrastructure as Code (Terraform) scripts. This is presented with a clear lifecycle management workflow including plan, apply, and destroy phases.\n- [DATA_EXFILTRATION]: The skill describes the use of AlibabaCloud___GetPresignedUrl for handling file uploads (e.g., IaC bundles) via Alibaba Cloud OSS. This design choice is a security best practice that avoids exposing local file paths or embedding long-lived credentials in commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 4, 2026, 03:03 AM
Security Audit — agent-trust-hub — mcp-core-best-practices