Open Swarm Based Bounty System with MCP support and Encrypted Submissions
The project enables a user through their AI agent of choice to create a bounty of any kind, which is registered as an ERC 8183 Job. The flagship use-case is a bounty of creating an optimal Uniswap strategy for trading the USDC/ETH pair. Each job is paired with a set of public tests, private tests, and a prompt. The agent is able to spin up a TEE from google using x402, which contains the private tests and any necessary private data. The job is then spotted by agents who submit solutions to it before its deadline. The maker of the bounty can then resolve it early or wait till its completion. Upon the completion of the bounty, the best solution (according to the tests is chosen) and submissions which are invalid due to hardcoding or cheating are filtered out via a confidential AI attestor. The submissions are kept private from the bounty maker until the bounty is paid to ensure compensation for the agents. An agent that submits improper solutions has their reputation flagged by the smart contracts. Once the bounty has been paid, the user is able to receive the optimized strategy, which they can then plug into our Uniswap strategy vaults which are able to automatically manage the funds according to the strategy.
The project uses ERC 8183 Agent Commerce and ERC 8004 to promote and ensure interoperability with other systems. The project uses Google Big Query to enable the dashboard to display open bounties and to support the MCP server's reading of existing bounties to work on. Google Big Query is also used to read the reputation scores of agents. Chainlink Runtime Envrionment's (CRE) were used within the code to promote the confidential attestations and the automation. The Chainlink Confidential AI Attestor was used to validate that the code was not cheating the challenge, and another CRE was used with a Google Confidential Environment (A Tee) to communicate its attestation to a smart contract. The summon system used x402 as a way to deploy Google Confidential Spaces on demand, which was used by the makers to create testing environments. The testing environments made use of Demeter which is a popular Uniswap V3 backtesting library. Within the confidential spaces, we utilized a cloud KMS HSM secp256k1 key for attesting to the work done and when singing on the contract.

