Verifying Non-Risk Status

I'm very interested in trying to find ways to distribute helping people. Not in micro-donations, but more that if there's someone in my neighborhood that is in need - the elderly, someone going through some home issues, lacking food - I'd love to be able to pull that up on my phone, see the needs and help where I could. Likewise, with enough scale, I think that a lot of simple acts of kindness could occur using the Internet for people to send alerts and then people to volunteer for quick jobs. Sort of like how you already most likely help your friends and family. 

However, it only takes one runaway to be assaulted or one soup kitchen to have massive amounts of food poisoning, and the whole things comes crashing down. While my original thoughts in this matter were toward the latter - how can I get it so in-home prepared food is safe for distribution - the former is much more serious than getting around the legal pains of the health department. 

There are a couple options to verify that we're working with people that have the best intentions and we can make sure they execute their help correctly.

First, is the system we have now, which is essentially people vouch for other people. This is what friends do, and friends of friends when you need help after your car breaks down. But how much vouch is needed and who provides it. Essentially, a seller rating system like on eBay, this doesn't mean that the vouchers don't also have a low bar of responsibility such that the quality of folks is inflated. This isn't necessarily meaning that anyone is doing anything malicious after all - it's just they may not know how to cook good meals for the elderly. 

The second is another system we have now, which I'll casually call the "church" system. We have some central authority that is essentially responsible for the activities of its members and they have to be added and approved by those members. I like this system a bit more, but if we're using say a cloud service to support these communities, how do we then go and vet those communities. This solutions seems to lean itself towards a more distributed service, where, if a "church" wanted to get this up and running they could, but whoever was writing the software would just sorta put it out there and hope for the best.

I'm okay with this idea, and it is making giving communities more efficient, but it's not doing what I hoped the system could do, which is to get folks everywhere to find small ways they could help their community with immediacy and in a less bounded context. Even if you're in a giving community, they may not be aware that someone 1 block from your house is in need, so there's some pre-/self-selection in those activities, which goes back to the point above, that's it's really about efficiency then. 

At this point I really don't have an answer, but I hope putting it out there, may, for the non-existent readers of this website inspire some things towards a safer and more assured way to help those in need.