Take a curve $y = f(x)$ and let $\psi(x)$ be the angle between the curve and the horizontal at the point $x$.
Letting $ds$ be an arc-length element (so $ds^2 = dx^2 + dy^2$) we have that $\tan \psi = \frac{dy}{dx} = y'$, $\sin \psi = \frac{y'}{\sqrt{1+(y')^2}} = \frac{dy}{ds}$ and $\cos \psi = \frac{1}{\sqrt{1+(y')^2}} = \frac{dx}{ds}.$
How can we show from this that $\kappa := \frac{d\psi}{ds} = \frac{y''}{\left( 1 + (y')^2\right)^{3/2}}$?
This claim is made in lecture notes I'm reading.
I've thought about using the idea that $\frac{d\psi}{ds} = \frac{\partial\psi}{\partial x} \frac{dx}{ds} + \frac{\partial\psi}{\partial y} \frac{dy}{ds}$ but can't make this work because $\psi$ does not have explicit dependence on $x$.
Similarly, writing $\psi = \arcsin (\dots)$ leads to the need to calculate $\frac{d}{ds} F(y')$, for some function $F$ of the derivative $y'$, which I don't know how to do.
Is there a more straightforward approach?

You have $\tan\psi=y'$ and therefore $(1+\tan^2\psi)\psi'=y''$, or $$\psi'={y''\over1+\tan^2\psi}={y''\over 1+y'^2}\ .$$ Now by the chain rule $${d\psi\over ds}={d\psi\over dx}\>{dx\over ds}={\psi'\over s'}={\psi'\over\sqrt{1+y'^2}}\ ,$$ so that $$\kappa={d\psi\over ds}={y''\over(1+y'^2)^{3/2}}\ .$$