sentinel-V4

Sentinel v4 is an agentic liquidity management protocol for Uniswap v4 that unifies an immutable hook (safety) with deterministic automation (execution) to keep LP capital productive. LPs deposit into a pool‑scoped share ledger, so each pool has isolated accounting of active liquidity, idle balances, and optional Aave v3 yield positions; the share price reflects total NAV, making exits and entries proportional and auditable. The hook enforces oracle‑based deviation checks on every swap, preventing manipulation and ensuring that rebalances happen within safe price bounds. An off‑chain agent continuously watches tick crossings, price drift, and utilization; once thresholds are met it calls maintain to withdraw active liquidity, recall idle funds from Aave, compute a new optimal range around the current price, and redeploy capital into the fresh band while re‑supplying any remaining idle balances. This maximizes yield by (1) keeping more liquidity in active fee‑earning ranges, (2) monetizing idle capital in Aave when it would otherwise sit unused, and (3) minimizing out‑of‑range time and operational lag. It “saves” LPs by actively managing positions against volatility, preventing stale ranges, reducing capital inefficiency, and enforcing transparent, rule‑based execution that limits trust assumptions and cross‑pool risk.

We built Sentinel v4 as an agent‑first liquidity protocol where the DON agent is the operational “brain.” It continuously watches tick crossings, price‑oracle deviations, and pool utilization, then deterministically calls maintain to withdraw current liquidity, compute a new centered range, and redeploy capital while recalling or re‑supplying idle Aave v3 balances. We initially planned to run this agent on Chainlink CRE, but access limitations forced a pivot; we also tested Chainlink Functions for agent compute, but token consumption was too high, so we consolidated on Chainlink Automation for reliable, low‑overhead execution. On‑chain, Solidity Uniswap v4 Hooks enforce immutable safety (oracle deviation checks on every swap), pool‑scoped accounting (PoolId isolation), and NAV‑based LP shares across active + idle + yield assets, with Aave v3 handling idle yield. We used Foundry for scripting, deployment, and unit/fuzz/integration tests; TypeScript/Node for deployment tooling and address/config updaters; and a Vite + React + TypeScript UI to visualize pools and LP state. This stack keeps the agent decisive, verifiable, and trust‑minimized while the hook guarantees safety and transparent accounting.