Potential fix for code scanning alert no. 2: Uncontrolled command line

[Rootless-Ghost]

Potential fix for https://github.com/Rootless-Ghost/AtomicLoop/security/code-scanning/2

General fix: do not accept raw executable command text from the API. Accept a stable identifier (for example command_id) and resolve it server-side to a hard-coded/embedded allowlisted command literal, then execute only that resolved literal.

Best fix here without changing core execution behavior: in app.py’s execute endpoint, ignore body["command"] for execution and require a command_id that maps to a server-side command dictionary. Keep command only for compatibility in response/logging if needed, but execution must use resolved command text from the allowlist mapping. In core/executor.py, keep existing _is_allowed_atomic_command check as defense in depth.

Concretely:

• File: app.py, in the execute route block (around lines 242–283).
• Add a small embedded mapping from IDs to commands (or a helper lookup if already present in shown snippet scope).
• Replace direct assignment command = str(body.get("command", "")).strip() with:
  • read command_id,
  • resolve command = COMMANDS.get(command_id),
  • reject if missing/unknown.
• Keep existing _is_allowed_atomic_command(command, executor_type) validation.
• No changes required in core/executor.py for sink call once untrusted raw command input is removed at source.

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