Command feedback is really insufficient

**I'm submitting a feature request**
- **Library Version:**
0.19.0

**Please tell us about your environment:**
- **Operating System:**
Windows 10
- **Node Version:**
6.3.0


- **NPM Version:**
3.10.6


- **Browser:**
N/A
- **Language:**
N/A

**Current behavior:**
All invalid cli commands return "invalid command: {command}"

e.g.

```
au run --watch
Invalid command: run

{Create new project text}
```

**Expected/desired behavior:**

- **What is the expected behavior?**
"Invalid command, unfulfilled dependencies. Run npm install"
- **What is the motivation / use case for changing the behavior?**
Any users working on different systems or for team-based projects, there is inevitably individuals who might forget to install dependencies or something along those lines. Significantly improving the output of `au {command}` would save a meaningful amount of time, especially considering the frequency this is likely to have occurred.