PatchFlow — Agentic Debugging Workflow

Software debugging is often fragmented across logs, code, tests, issue reports, and manual investigation. Developers may spend significant time identifying the root cause, deciding what to change, and verifying whether a fix actually resolves the problem. PatchFlow is an interactive agentic debugging workflow designed to bring these stages together into one structured process. PatchFlow takes a software incident and guides it through investigation, root-cause analysis, patch generation, validation, impact analysis, and reporting. Instead of treating debugging as isolated tasks, the workflow organizes the information and actions into a traceable sequence so that developers can understand not only what was fixed, but also why the fix was proposed and how the result was validated. The prototype is demonstrated through an OrderFlow API incident involving a checkout total mismatch. The user can move through the complete debugging lifecycle using dedicated workspace sections for Overview, Investigation, Root Cause, Patch, Validation, Impact, and Reports. Supporting sections for Evidence and Test Suite provide additional context and verification information. The target users are software developers and engineering teams who spend time investigating bugs, understanding unfamiliar codebases, applying fixes, and validating changes. PatchFlow is designed to reduce fragmented manual work and make debugging more structured, explainable, and easier to review. What makes PatchFlow distinctive is the way the debugging process is presented as an end-to-end workflow rather than as a standalone AI chat interface. Each stage contributes to the next, creating a clear chain from incident to resolution and making the debugging process easier to follow and audit.