[guided]The object to construct is not an arbitrary Ricci flow; it is the specific complete rotationally symmetric flow starting from the standard capped-cylinder initial metric. Let $M$ be the smooth manifold underlying this capped-cylinder model. Because the model is a cylinder with one smooth cap, it is diffeomorphic to $\mathbb{R}^3$; fix a diffeomorphism $\Phi: \mathbb{R}^3 \to M$. This is the point at which the statement's dimension-three ambient space is matched with the manifold used in the construction.
Let $g_0$ be the standard capped-cylinder initial metric on $M$. The hypotheses of Perelman's standard-solution existence theorem are met: $g_0$ is smooth, complete, rotationally symmetric, has positive curvature operator, agrees with the standard round shrinking cylinder outside the cap region, and has bounded curvature together with bounded covariant derivatives of curvature at time $0$. The theorem produces a smooth map $g: M \times [0,1) \to \operatorname{Sym}^2(T^*M)$, where $g(p,t)=g(t)_p$ and $\operatorname{Sym}^2(T^*M)$ denotes the bundle of symmetric covariant $2$-tensors on $M$. It satisfies
\begin{align*}
\frac{\partial g}{\partial t}(t) = -2\operatorname{Ric}_{g(t)}
\end{align*}
for every $0 \leq t < 1$, where $\operatorname{Ric}_{g(t)}$ is the Ricci curvature tensor of $g(t)$. The theorem also gives completeness of $(M,g(t))$, rotational symmetry, and the normalisation of the maximal time interval to $[0,1)$.
Now pull the solution back by $\Phi$. Since pullback by a diffeomorphism preserves the Ricci-flow equation, completeness, rotational symmetry, and curvature positivity, $(\mathbb{R}^3,\Phi^*g(t))_{0 \leq t < 1}$ is a standard solution in dimension three. Perelman's theorem states that the curvature operator is positive for each $t \in (0,1)$, so the curvature remains positive on the required positive-time interval. Finally, the compact-time estimates supplied in the standard-solution theorem give, for every $T \in [0,1)$ and every integer $m \geq 0$, a finite constant depending only on $T$ and $m$ that bounds the $m$-th covariant derivative of curvature on $\mathbb{R}^3 \times [0,T]$. These bounds include the endpoint $t=0$ because the initial capped-cylinder metric has bounded curvature derivatives; Shi's derivative estimates explain the propagation of such local derivative control for positive time, but the standard-solution construction is the cited source for the uniform compact-time statement including $t=0$.[/guided]