Nothing to hand? Load the — a New message button that opens the composer and closes it again in the same click, a Save draft that never saves, a counter that advances by one instead of two, two refresh requests fighting over one flag, a Send guarded by state it just cleared, and an effect that undoes an edit — or the , where the correct verdict is clean and the useful output is which store actions are still risky for the next caller.
Paste the components and the store — the map is free
No upload, no AI: the prescan reads your source in the browser and builds the thing this bug class is invisible without — the store side-effect map. Every action, the fields it sets, the fields it quietly clears, and which of those belong to a different action. Then it traces each touchpoint's handler and runs nine deterministic checks: sequential undo, dangerous resets, async races, stale closures, Save-that-never-saves, work behind an unreachable guard, effects that undo the click, non-atomic multi-store writes, and controls with no handler at all. Each group explains why it matters. This costs nothing and happens while you type.
The AI traces every path — this is the metered part
The model reads what the prescan cannot judge: whether the final state actually matches what each control's own label promises. It returns the side-effect map with ownership and risk, a per-touchpoint trace table with the conflicting step marked, and one finding per real bug — pattern, expected state, actual state, the mechanism and a corrected fragment. It may only cite touchpoints, actions and fields that appear in your paste, and every prescan flag is confirmed or explicitly set aside. Pricing is honest: a worst-case amount is reserved before the run and only what the run uses is charged — the meter next to the button shows both.
Fix, export, re-audit
Every corrected snippet in one paste-ready block, the findings as GitHub PR review comments with ```suggestion fences, a tickable checklist for the ticket, the store map as Markdown or CSV to keep next to the store itself, the traces as CSV, and Markdown or JSON of the whole audit. Then start fixing: the prescan re-runs as you edit and the strip above the run button counts the flags you have cleared, the ones still open and any you have just introduced — in the browser, for free, before you pay for a second audit. Audits are saved to your SkillSafe account when you are signed in, so they follow you to another machine, and restore puts the code back in the form too.
Derived from the @affaan-m/click-path-audit skill.