What problem does it solve? When a bug is reported, the default failure mode is to read a stack trace and propose several confident but wrong fixes in a row, each introducing new bugs. This Skill forces a disciplined debugging sequence so the root cause is confirmed before any fix is written, which matters most when bugs span multiple layers like Firestore security rules, an API adapter, and DOM state. ## Core Features & Use Cases - Six-step enforced sequence: Reproduce, Minimize, Hypothesize, Instrument, Fix, and Manual Test, with an explicit gate that forbids proposing fixes before hypotheses are stated. - Multi-layer hypothesis framing: Generates 2-3 hypotheses that each name a specific layer and mechanism, such as Firestore rules denials, un-awaited batch writes, or optimistic DOM updates masking failed writes. - Discriminating instrumentation: Picks one cheap test (a targeted console.log, network tab inspection, Firestore console check, or breakpoint) instead of scattering noisy logging. - Use Case: A user reports duplicate papers appearing after import. The Skill walks through exact repro steps, narrows to the minimal triggering input, hypothesizes whether the adapter's batch write or the rules layer is at fault, confirms with one instrumented check, then delivers a minimal fix with a manual verification test and changelog reminder. ## Quick Start Ask the assistant to run the diagnose skill on a bug you are seeing, for example by describing the unexpected behavior and the steps that trigger it.
