Interoperable FROST co-signing for shielded ZEC: open specs and a reference implementation

Hi @Khadim, congrats on #1432, and nice work on the padding check that got fixed in Zkool.

I’m Babacar, founder of lomi., a payment company in Abidjan. You wrote that a ZIP for the co-signing messages makes more sense once two wallets use them. We’d like to be one of them.

We build Nomad, an assistant people let act and pay for them. We wanted its spending limits to hold even if the agent’s machine gets taken over. So the key is split 2-of-3: the agent holds one share, a policy signer holds the second, and the user holds the third. The policy signer reads the PCZT itself and only signs inside the limits the user set. Above a limit it doesn’t sign and asks the user instead, and then only the user’s share can sign with the agent.

It works on mainnet:

All three shares are still on one machine, so this proves the signing rules, not the separation. Three devices is next. We’re on the same stack as you and Zafe (frost-core 3, reddsa 0.5.2). Code: GitHub - lomiafrica/cosigner: FROST 2-of-3 co-signer so a Zcash agent cannot outspend its mandate. · GitHub

Three things we’d like to do with you, if they help:

  1. Join the interop test you’re planning with Zafe after NU7: your co-signer and ours signing the same PCZT.
  2. Export the user share as the zkfrost string from #1432, and tell you anything your vectors catch.
  3. Write the policy signer’s part of the co-signing messages (allow, ask the user, refuse, and the rule that it reads the sighash and alphas from the PCZT, never from the coordinator), as a section of your spec or a separate ZIP that cites it, whichever you prefer.

If any of this overlaps with what you already planned, tell me and we’ll work around it. Thanks!