Show HN: A hook that stops Claude Code from re-trying fixes that failed (github.com)

A new tool called RegressionLedger has been introduced to improve the functionality of coding agents like Claude Code by preventing them from repeatedly applying failed fixes. This tool addresses a common issue where coding agents lose memory of previous attempts when the context window resets, leading to a frustrating loop of repeated errors. RegressionLedger works by fingerprinting each edit made by the agent, linking it to the outcome of subsequent tests (pass or fail), and maintaining a ledger that survives session restarts. If the agent attempts to reapply a previously failed fix, RegressionLedger blocks the action and provides the reason, prompting the agent to try a different strategy. This innovation is significant for the AI/ML community as it targets a critical pain point in the development cycle involving AI coding agents by offering a structured approach to prevent redundant attempts and conserve resources. With a zero-dependency design, easy setup, and local storage of data, RegressionLedger enhances the reliability of coding agents while offering developers the ability to audit past attempts through a user-friendly command-line interface. By improving the way these agents learn from mistakes, RegressionLedger can lead to more efficient workflows and better debugging outcomes.