safe-agent-executor

CLI Agent Sandbox Guard: A zero-dependency safety wrapper to execute terminal commands under validation, checking for malicious/destructive code patterns and setting strict execution limits.

⚡ Features

• Zero Runtime Dependencies: Fast and secure native command execution.
• Dangerous Commands Shield: Statically parses and blocks destructive commands like recursive deletes (rm -rf *), write attempts to system sectors, and git force-pushes.
• Execution Limits: Enforces customizable execution timeouts (in milliseconds) and terminates hanging processes cleanly.
• Node spawn Wrapper: Forwards child output stream buffers and exit code signals accurately.

📦 Installation

npm install safe-agent-executor

🚀 Usage

1. CLI Wrapper

Run a command through the safe sandbox:

npx safe-agent-executor "echo 'hello'"

# Blocking destructive commands
npx safe-agent-executor "rm -rf *"
# Output: [Security Violation] Dangerous command blocked by safety policy.

# Force allow dangerous operations
npx safe-agent-executor "git push origin main --force" --allow-dangerous

# Enforcing a 500ms timeout
npx safe-agent-executor "sleep 2" --timeout 500

2. Programmatic API

import { runCommand } from 'safe-agent-executor';

try {
  const result = await runCommand("npm run dev", [], {
    timeout: 5000 // limit to 5 seconds
  });
  
  if (result.timedOut) {
    console.log("Process timed out and was terminated.");
  } else {
    console.log("Output:", result.stdout);
  }
} catch (err) {
  console.error("Execution blocked:", err.message);
}

📄 License

MIT License.