Descripción
Tokenized stocks can't become useful DeFi collateral until someone prices their risk on-chain. Sigma does exactly that. A parametric portfolio Value-at-Risk engine runs in Arbitrum Stylus (Rust → WASM), produces VaR identical to a Solidity baseline deployed alongside it for verification, and is called by the Sigma Vault on every borrow and withdrawal. The non-custodial Vault accepts the five live Robinhood Chain stock tokens (TSLA, AMD, AMZN, NFLX, PLTR), lends real testnet USDC against the basket, and bounds leverage with a correlation-aware VaR plus an independent 80% base-LTV backstop. A separate Strategist contract enforces user-signed policies before any AI-agent proposal can touch a position — and every executed action emits its rationale on-chain, so the audit log is the chain. Prices come from a cross-checked Pyth + RedStone oracle. Built for the Robinhood Chain reserved slot and the Best Agentic Project track.
Progreso del hackathon
Sigma was designed and built from scratch during the buildathon — there was no pre-existing codebase.
- Day-1 feasibility spike: confirmed Stylus deploys and executes on Robinhood Chain testnet (chain 46630), verified the RPC, faucet (ETH + stock tokens), and Pyth/RedStone price availability de-risking the whole architecture before writing product code.
- Core risk engine: built the parametric portfolio VaR computation in Rust/Stylus (WAD fixed-point, signed correlations, Babylonian sqrt, packed correlation matrix), then wrote an equivalently-specced Solidity baseline and proved both produce identical VaR on the deployed reference vector ; so the headline gas comparison (Stylus 122,318 vs Solidity 128,960 gas, a 5.15% saving at block 70,915,096) is reproducible, not asserted.
- Vault + Strategist: shipped the non-custodial lending Vault (VaR-bounded borrows, 80% LTV backstop, liquidation with close-factor and bonus caps) and the agent-bound Strategist (per-user policy: agent address, borrow cap, concentration limit, min health, cooldown), backed by 40 Foundry tests including 512-run fuzz invariants and Slither static analysis with 0 findings.
- Pilot agent + dashboard: built the policy-bound LLM agent (tool-using, provider-agnostic) and a live Next.js dashboard reading every number directly from the deployed contracts.
- Hardening sprint (final days): ran a full internal security review plus an external audit, remediated every finding, and migrated to a fresh Vault + Strategist on testnet. Added oracle-updater resilience (retries + per-symbol fault isolation), a borrow-liquidity clamp, an on-chain stress lab (scenario VaR computed by the live Stylus core), a first-fold proof strip, and an owner-powers disclosure.
- Live & verifiable: deployed app at https://sigma-two-iota.vercel.app, all contracts verified on Robinhood Chain testnet, CI green, with deployment and audit records committed to the repo.
Estado de recaudación de fondos
Sigma is a pre-product, testnet-only build with no external funding, no token, and no raise to date; all development was bootstrapped during the buildathon. The contracts are deliberately framed as a "risk engine + proof Vault," not a production lending market, and the limits to mainnet (independent professional audit, multisig + timelock on owner powers, real lender accounting, decentralized oracle) are documented openly. Near-term, we'd pursue ecosystem/RWA grants (e.g. the Arbitrum Foundation / Robinhood Chain path) to fund a professional audit and a production lending integration before any capital raise.
