What problem does it solve? AI agents and developers frequently misintegrate Pyth oracles on EVM chains: exposing updatePriceFeeds as a standalone function (enabling sandwich attacks), hardcoding update fees, using wrong per-chain contract addresses, mishandling negative price exponents, and ignoring confidence intervals. This Skill provides correct, production-grade integration patterns for Pyth's pull-based oracle model. ## Core Features & Use Cases - Anti-Sandwich Integration Patterns: Atomic updatePriceFeeds + price read in a single transaction, with dynamic fee computation via getUpdateFee and confidence interval validation. - Hermes API & TypeScript Client: Fetch latest or historical price updates from Hermes, stream prices via SSE, and submit updates on-chain using viem. - Reference Data: Per-chain Pyth contract addresses, bytes32 feed IDs for major pairs, error code tables with fixes, and gas cost benchmarks. - Advanced Flows: Sponsored feed fallback logic, historical benchmark prices via parsePriceFeedUpdates, multi-feed batch updates, and Express Relay integration for MEV-protected liquidations. - Use Case: Build a lending protocol on Arbitrum that liquidates undercollateralized positions using fresh ETH/USD and BTC/USD prices, with confidence validation and MEV-protected liquidation routing through Express Relay. ## Quick Start Ask the agent to write a Solidity contract that consumes the Pyth ETH/USD price feed on Arbitrum using the atomic update-and-read pattern with confidence validation.