$x^2 + bx$
$=x^2 + bx + c - c$
$=(x + k)^2 - c$
$=x^2 + 2kx + (k^2 - c) = x^2 + bx + 0$
This implies:
$2k = b$, so $k = b/2$, and:
$k^2 - c = 0$, or $k^2 = c$, or $(b/2)^2 = c$
So to complete the square we are making the transformation:
$x^2 + bx \implies (x + b/2)^2 - (b/2)^2$
Your approach is just fine. This is often how mathematics is done; in the initial stages, you may run into a dead end or use unnecessary variables. Later on, once you have found that $k$ does exist and equals $b/2$, you can rework your exposition to avoid $k$ entirely and just use $b/2$. You will often see proofs in textbooks where the author "magically" knows what substitutions to use, etc.; this comes from experience and from working through the problem as you have done, then presenting it as if it were obvious all along.
That said, there's a circularity in your first few lines that is confusing and should be avoided. You wrote:
It's strange to begin and end with $x^2+bx$, since it isn't clear where you're going. Instead, leave off that last line, thus: