sakura-fix

What problem does it solve? When a test, build, runtime, or integration check fails, teams often repeat the same unsuccessful patch or lose track of what was already tried. This Skill turns an observed failure into an evidence-supported root cause, the smallest corrective change, and a successful rerun of the original check, while preserving the full repair history across conversations. ## Core Features & Use Cases - Evidence-Based Diagnosis: Reproduce the failure, compare expected versus actual behavior, and form one root-cause explanation that current evidence can support or reject before changing any code. - Bounded Repair with History: Apply the smallest fix without disabling validation or weakening assertions, then rerun the original failed check and the closest affected regressions. Failed attempts and hypotheses are recorded so a new conversation never resets the repair history. - Escalation and Recovery Rules: After two failed repairs, stop that direction, preserve the reproduction, and move diagnosis to a fresh context or a materially different recovery path such as replacing the implementation approach. - Use Case: A CI test fails after a refactor. Use this Skill to attach the failure to the affected work item, diagnose the first meaningful divergence, apply a minimal fix, rerun the failing test, and record the version-increment event for the completed bug fix. ## Quick Start Use the sakura-fix skill to diagnose this failing test, find the root cause, apply the smallest fix, and rerun the original check.