Potential fix for code scanning alert no. 15: Uncontrolled command line

[whyseu]

Potential fix for https://github.com/agent-network-protocol/anp-agent-openchat/security/code-scanning/15

The best way to fix this problem is to implement an allowlist of acceptable commands. This ensures that only predefined, safe commands can be executed. We will define a dictionary of allowed commands (like the example provided in the background section) and validate the user input against this allowlist.

Specifically:

1. Introduce a dictionary ALLOWED_COMMANDS mapping valid user inputs to their corresponding commands.
2. Modify the start_instance function to verify that the command argument is in the ALLOWED_COMMANDS dictionary. If not, raise an exception and avoid executing the command.
3. Update the REST API endpoint /api/instances to reflect this validation.

---

Suggested fixes powered by Copilot Autofix. Review carefully before merging.