I think emulating existing darknet market infrastructure makes sense to address a couple of the issues you mentioned- like public reviews & messaging for employment + moderation & escrow/multisig for transactions.

+ your third point, I think another issue is basically verification: what's stopping cheap, non-refugee laborers from scooping job postings from refugees? I don't think it would be possible to verify the refugee status of someone without leaking identifying info.

I guess your last suggestion sounds like the best thing to do.