We got tired of waking up to noisy CI failures and flaky tests hiding the real issues, so we built Claude Code Watchdog. It's a GitHub Action that automatically analyzes test failures in your CI pipeline, classifies them based on severity, and provides intelligent fixes or detailed GitHub issues to help your tests become self-healing. How it works: * Analyzes your last 20 workflow runs to distinguish chronic failures from occasional flakes. * Classifies tests by severity (critical, frequent, intermittent, isolated). * Creates detailed, actionable GitHub issues with context and recommended fixes. * Automatically fixes straightforward issues by opening PRs when confident. Costs about $0.20 per failure analyzed (via Anthropic API). We've been using it internally for API monitoring and integration tests, significantly reducing noise and helping us catch critical problems quickly. |