What problem does it solve? Runtime errors, test failures, and build issues in a Spring WebFlux backend pipeline often get fixed through unbounded trial-and-error that leaves partial fixes stacked on top of each other. This Skill enforces a structured debugging session that reproduces the problem, tests at most three hypotheses with full reverts between attempts, and records an auditable trail. ## Core Features & Use Cases - Bounded hypothesis testing: A debugger agent tests a maximum of 3 hypotheses, fully reverting each failed attempt before trying the next, and never modifies tests or adds suppressions to silence failures. - Pipeline state integration: Acquires a lock on the feature's progress file, records classification, root cause, and modified files, then suggests the correct pipeline re-entry point (re-verify, re-review, or resume implementation). - Evidence-backed reporting: Produces a debug report where every hypothesis carries observed evidence and a file:line location, with an escalation path when all hypotheses fail. - Use Case: A NullPointerException appears in EmployeeHandler.createEmployee. Invoke the skill with the error message; it reproduces the failure with a WebTestClient case, applies a null-check fix in the validator, runs the full regression suite, and reports 47/47 tests passing. ## Quick Start Ask the AI to debug the pasted error message or failing test, for example by running be-debug with "NullPointerException in EmployeeHandler.createEmployee at line 42".
