>ju'i mi cusku sa'u le nu
>

I suggest {do'u} after {ju'i}, else {ju'i mi} forms a single free modifier, which I highly suspect is unintentional.
Furthermore, cusku2 meant to be text (produced either by quotation, the {se du'u} construct, or description of another place that is meant to be text.) {le nu ...} is an event, not text, and is therefore rather odd in this location. (I say "odd" rather than "wrong" since one can still understand what you mean (at least I think.))

>mi puba'o citka
>le so'o zirsmela
>noi nenri
>le lektanxe ku'o
>
>zi'e noi
>do puca'o lakne
>ralte
>fi'o tezu'e cersai
>

I'm guessing this is meant to translate "have for breakfast". In any case, {fi'o tezu'e cersai} doesn't parse, and {ralte} isn't quite correct for that sense of "have", which really just means eat. "I'll have a coffee," just means that you're going to *drink* the coffee. I suggest using the modifier {la'a} rather than actually invoking {lakne} on the selbri, but the reasons are a bit too complicated to explain at this point.
All in all, I'd suggest {zi'e noi / la'a do pu ca'o / citka / co cersai}.

>.i .u'ucui ko fraxu mi go'i

{go'i} is a selbri, and you'd need to encapsulate it into an abstraction (in particular, an event) in order to make it fill fraxu3: {ko fraxu mi lo nu go'i}.
(And typo {cui} -> {cu'i}.)

>.i be'unai oinairo'o da pu kukte

If you're going to put a dot before "u'u", you might as well put one before "oi" too :) .
{da} is not anaphoric, by all means; you can't use it to refer back to something already said. In this context, I'd suggest just leaving out the pronoun altogether. The only this in context for which it would make sense to be in kukte1 are the {zirsmela}.

>.i mutce titla .e da mutce lenku

{.e} is used to connect sumti (and numbers), whereas here, once we drop the {da} for the same reasons as above, we need to connect bridi-tails. This is done with {gi'e}:
{.i mutce titla gi'e mutce lenku}
However, there's some redundance (which very well may be intentional, since it's poetry.
A common error when using {mutce} is to use it as the modifier. This makes sense due to naïve indo-european natlang assumptions, but is somewhat wrong, overall. Consider how you would phrase it as a gismu deep structure (a type of bridi consisting only of primitive selbri, (e.g. not tanru,) abstractions, and description sumti.). You'd say {.i mutce lo ka/ni lenku}. Therefore, the actual main claim is one of {mutce} and not one of {lenku}. Of course, {mutce lo ka/ni broda} implies {broda}, so the {lenku} claim is still made.
Thus, {.i titla mutce gi'e lenku mutce}. We can further simplify by using a tanru-internal AND connective {je}: {.i titla je lenku mutce}.