ToolTrust
D63/100
elusznik

mcp-server-code-execution-mode

mcpmain@27d23b8e2c76

@elusznik

An MCP server that executes Python code in isolated rootless containers with optional MCP server proxying. Implementation of Anthropic's and Cloudflare's ideas for reducing MCP tool definitions context bloat.

By elusznik | 17 findings | Scanned 6/3/2026 | tooltrust-scanner/v0.3.13 | 333

2 Critical4 High5 Medium1 Low5 Info

Risk Summary

Block in Production

Excessive Permissions + Dep Visibility risk is significant. Avoid using this in production agents.

Potential impact: The agent may gain overly broad access to files, network, databases, or execution capabilities.

Recommended action: This tool should stay disabled in production agents until the flagged risks are fixed and the scan is clean.

{
  "mcpServers": {
    "mcp-server-code-execution-mode": {
      "disabled": true
    }
  }
}

Security Findings (17)

  • CriticalAS-006

    💻Arbitrary Code Execution ×2

    heuristic signal

    tool name or description implies arbitrary script/code execution (evaluate_script, execute javascript, etc.)

    CODEINTERPRETER_CREATE_SANDBOXCODEINTERPRETER_EXECUTE_CODE

    Fix: This tool can execute arbitrary code or shell commands on the host system. Remove it unless strictly required. If kept: (1) restrict access to trusted users/agents only, (2) require human approval before each invocation (Claude Desktop: set approval_required: true; other clients: enable equivalent confirmation), (3) use the most restrictive sandbox or read-only mode available, and (4) never expose this tool to untrusted input sources.

  • HighAS-002

    ⚠️Excessive Permissions ×4

    tool declares exec permission

    CODEINTERPRETER_CREATE_SANDBOXCODEINTERPRETER_EXECUTE_CODECODEINTERPRETER_RUN_TERMINAL_CMD

    tool declares network permission

    CODEINTERPRETER_EXECUTE_CODE

    Fix: Tool requests broad permissions (exec/fs/network). Validate input parameters using Enums where possible, and restrict file system operations to explicit allowed directories.

  • MediumAS-002

    ⚠️Excessive Permissions ×5

    tool declares fs permission

    CODEINTERPRETER_CREATE_SANDBOXCODEINTERPRETER_EXECUTE_CODECODEINTERPRETER_GET_FILE_CMDCODEINTERPRETER_RUN_TERMINAL_CMDCODEINTERPRETER_UPLOAD_FILE_CMD

    Fix: Tool requests broad permissions (exec/fs/network). Validate input parameters using Enums where possible, and restrict file system operations to explicit allowed directories.

  • LowAS-011

    ℹ️Missing Rate-Limit / Timeout

    CODEINTERPRETER_CREATE_SANDBOX:tool performs network or execution operations but declares no rate-limit, timeout, or retry configuration

    Fix: Declare explicit rate-limit, timeout, and retry configuration for all network and execution tools. Implement exponential back-off and surface resource state to the calling agent.

  • InfoAS-014

    ℹ️Dependency Inventory Unavailable ×5

    Tool did not expose metadata.dependencies or repo_url, so supply-chain coverage is limited.

    CODEINTERPRETER_CREATE_SANDBOXCODEINTERPRETER_EXECUTE_CODECODEINTERPRETER_GET_FILE_CMDCODEINTERPRETER_RUN_TERMINAL_CMDCODEINTERPRETER_UPLOAD_FILE_CMD

    Fix: Review and remediate the identified issue.

Scan this tool yourself

Reproduce this audit locally, integrate into CI, or let your agent audit its own tools.

Install once, then scan any MCP server:

$ curl -sfL https://raw.githubusercontent.com/AgentSafe-AI/tooltrust-scanner/main/install.sh | bash
$ tooltrust-scanner scan --server "npx -y mcp-server-code-execution-mode"

Adjust the package name if your npm registry name differs from the tool ID. View source

Add badge to your README

Copy this Markdown to show your ToolTrust grade on GitHub.

[![ToolTrust Grade D](https://raw.githubusercontent.com/AgentSafe-AI/tooltrust-directory/main/docs/badges/grade-d.svg)](https://github.com/AgentSafe-AI/tooltrust-directory)