Let $H \rtimes K$ be the semidirect product of groups $H$ and $K$, not necessarily subgroups of anything.
In Theorem 10, part (5) of Dummit & Foote, Algebra, there's
Identifying $H$ and $K$ with their isomorphic copies in $H\rtimes K$, namely $\{(h,1)\}$ and $\{(1,k)\}$, we have
E) for all $h \in H$, $k \in K$, $k h k^{-1} = k \cdot h = \varphi(k)(h)$.
In other words, the hom into $Aut(H)$ associated with this semidirect product, $\varphi$, sends $k$ to conjugation by $k$ in $H \rtimes K$. Or something like that...
Their proof is:
$$ (1, k) (h, 1) (1, k)^{-1} = \\ (1 k\cdot h, k) (k^{-1} \cdot 1, k^{-1}) = \\ (k\cdot h, k)(1, k^{-1}) = \\ (k\cdot h, 1) $$
$(k\cdot h, 1) = (\varphi(k)(h), 1) = \tilde{\varphi}(k)(h)$
This is all very confusing. Is there a way to clear things up?
I'm not getting how you can mix the two isomorphisms and $\varphi$
By definition, we have that for $\;a,c\in H\,,\,\,b,d\in K\;$
$$(a,b)(c,d):=(a\cdot c^b,bd)$$
Note that we usually denote $\;\phi(c)(b)=:b^c\;$ and then clearly we get conjugation in the semidirect product.
Now, first
$$(h,k)\in H\rtimes K\implies (h,k)^{-1}=\left(h^{-k^{-1}}\,,\,k^{-1}\right)\;,\;\;\text{since}$$
$$(h,k)\left(h^{-k^{-1}},k^{-1}\right)=\left(h\cdot\left(h^{-k^{-1}}\right)^k\,,\,kk^{-1}\right)=(h\cdot h^{-k^{-1}k},1)=(1,1)$$
So that now
$$(1,k)(h,1)(1,k)^{-1}=(1\cdot h^k,k)(1,k^{-1})=(h^k,1)\in H\times 1$$