# Composite Agent Trust Scores (CATS): Locking up Coins for Composite Trust DB Scoring Agent Spectrum to N-degrees Social Graph

Composite Agent Trust Scores is a system that locks up ai16z, and other eco-system coins to establish a Web of bonded trust scores for eco-system participants. It works in conjuction with the [Eliza Agent Registry](https://hackmd.io/GpTCmB7IR0WQocKsZaYsrA) if the agents lacks a token or is a bad actor.

## Why? Benefits for Agents & People

Allows new agents and human newbies to quicky bootstrap a network of agents that they can trust for transactions (like avoiding rugs or trusting a real robot agent inside your house). Agents benefit from trusted introductions to new users who will get some assurance from social graph that the agents are "good". Sort of like the concept of Apple's AppStore but with a decentralized gating mechanism that puts agents on a spectrum from bad to good.

People & Agents who invest more into rating other agents and correctly predict crowd sentiment around other agents trusts scores can achieve a very high composite trust score across the entire network.

Naturally anyone should gain advantages from having higher trust scores, just like in the real world.

## Monetization of the Eliza Trust Score NFT's

### Composite Agent Trust Scores (CATS)

Everytime someone in the network queries their WoT/Social Graph up to N degrees to gain a composite agent trust score (CATS) they pay a small fee which is partially redistributed to participants (network data validators and the agents the report scores to their NFTs), and part sent to the DAO. The more degrees in the WoT involved the higher the fees.

**Technical Requirements**: this may require some sort of L0(?) network thats capable of gathering data from the various L1's that Eliza agents publish these trust scores to via NFTs.

**Optional:** the network may use ZK tech so that the aggragated trust scores are online on the network but are encrypted so that only the nodes in the network can access and process the data; in this case the NFTs should have the trust score and conviction numbers encrypted as metadata to the public key of one or more of the network nodes

### Use case example

Alice (Addresses A1, A2, A3) is selling some NFT from her new Project X, Bob (Address B) is interested in buying Alice's new NFT, but he doesn't know Alice and is worried that she might rug Project X. Bob has created CATS NFTs signafying trusted relationships with Chip, Doug, and Eliza, none of whom know Alice (and therefor have not published a trust score for Alice), but some of their friends, Fran, George, Harry who they have signed various levels of trust relationships with all have published C for Alice; so in this case as they are within 2 degrees of Bob, he can ask the network to calc a trust score for Alice using ZK so we dont know individual trust scores of Fran, George, Harry, but can know some composite score which Bob can use to decide if he wants to transact with Alice

## Implementaton Option A: Using Secure Trusted Ageent Messaging Protocol

see [Secure Trusted Agent Messaging Protocol (STAMP) & Eliza Agent Registry](https://hackmd.io/@XR/ai16z-tokenomics/%2FGpTCmB7IR0WQocKsZaYsrA)

Basically, registered agents, send msgs to the network with trust scores of other agents they identify (can be registered or not)

## Implementaton Option B: Eliza Trust Score 3-way NFTs
We leverage the TrustDB system (would need to be expanded beyond just pricing/trading type things) and create a Web of Trust type network (not the same but inspired by [WoT](ttps://en.wikipedia.org/wiki/Web_of_trust)) linking agents, and their tokens together into a social graph that has individual trust scores.

Agent locks up some ai16z and other tokens in an NFT, potential tokens involved:

A. reported agent's token (amount = 0 means lowest trust score)
B. reporting agent's token (higher amount = higher conviction score)
C. 0.69 of $ai16z token (we make use of the token2022 standard memo field to include registered agent "keyid" per [Eliza Agent Registry](https://hackmd.io/GpTCmB7IR0WQocKsZaYsrA) and some other metadata)
D. If no (A) or (B) then just (C) will need to be a higher amount and include trust score and convition data in Memo field.

The agent can unwind the NFT which results in them getting the coins back respectively, but the 0.69 of $ai16z is sent to the DAO.

Alternatively, trust score and conviction score could be encrypted metadata using a public key(s) of some network validator(s).

Agents can adjust the amounts of tokens in the NFT in order to update the Eliza Trust Score or adjust the encrypted metadata in the NFT.

### Rehypothecating the ETS Nfts

This requires more research but perhaps something similar to CRV it may be possible to loan out the underlying tokens in these NFTs and allow Agents to earn yield, the DAO taking a small cut.