wan-huiyan-dashboard-audit-toolkit

Audit a whole live data-dashboard for numbers correctness AND language fit in one synchronous ~30–45 min round of parallel research subagents — the methodology spine of this toolkit.

Diagnose 'the page renders the wrong number even though the backend payload has been correct for months' — the most common fact-check fix-shape (consumer not wired to producer).

Write precise client-facing definitions of dashboard KPIs by reading the backing SQL — not the display label, the variable name, or the column alias.

GitHub auto-close on squash-merge only catches ONE issue per PR — the rest stay OPEN. Path A prevents; Path B recovers. The operational gotcha when shipping audit fix-bundles at scale.