What problem does it solve? Multi-step flows like onboarding, checkout, and wizards often hide defects: states users cannot escape, steps that can be skipped by calling endpoints directly, and state corruption caused by refresh or the back button. This Skill gives an agent a systematic method to model any flow as a state machine and hunt for these structural bugs. ## Core Features & Use Cases - State Machine Modeling: Maps any flow as entry → preconditions → action → state change → feedback → next state, labeling where each state lives (server, client, URL, cache). - Defect Detection Patterns: Tests for dead ends, skippable preconditions, impossible states, refresh/back-button desyncs, duplicate operations, and stale feedback. - Confidence-Graded Findings: Reports issues with evidence levels from CONFIRMED to SPECULATIVE, plus false-positive filters to avoid noise. - Use Case: Auditing a checkout flow to verify that calling the final payment endpoint directly without completing prior steps is rejected server-side, and that pressing back after submission cannot create a duplicate order. ## Quick Start Audit the onboarding flow in this repository and report any dead ends, skippable steps, or refresh and back-button problems as a state machine diagram with findings.
