What problem does it solve? Verifying that a new, upgraded, or resynced Ethereum node returns the same JSON-RPC responses as a trusted baseline is hard: raw diffs are flooded with benign serialization, tolerance, configuration, and sync-gap noise. This Skill drives the repo's runner compare tool to run a controlled cross-node correctness comparison and produce a findings report where only genuine defects survive. ## Core Features & Use Cases - Pinned-block comparison: Probes both endpoints, verifies chainId and block-hash agreement, and pins all calls to a static historical block so head drift never pollutes results. - Expected-difference rules: Suppresses benign version, numeric-tolerance, and error-wording differences via a YAML rules file, and auto-classifies configuration/capability errors (namespace disabled, range caps, pruned state) separately from real defects. - Categorized findings report: Writes comparison-results.json, an HTML report, provenance, and an ANALYSIS.md that separates real defects (with reproduction curls) from config deltas. - Use Case: After upgrading a Nethermind archive node, compare it against the old Geth baseline over a sampled corpus of RPC calls to confirm no response regressions before switching traffic. ## Quick Start Ask the AI to compare your baseline and candidate Ethereum RPC endpoints for response correctness using the compare-nodes workflow, providing both endpoint URLs.
