Is there a sequence of base-$b$ digits of length greater than one with all digits $\ne 0$ that does not represent a prime number in any base?
Example: $12_{10}=12$ is not prime, but $12_3=5$ is.
Is there a sequence of base-$b$ digits of length greater than one with all digits $\ne 0$ that does not represent a prime number in any base?
Example: $12_{10}=12$ is not prime, but $12_3=5$ is.
On
Extending Jorge's example, $$\sum_{k=0}^{n}{\binom{n}{k}b^k}=(b+1)^n$$ is not prime in any base large enough to have the binomial coefficients as digits.
On
How about $121$. We have $1b^2+2b+1=(b+1)^2$.
Similarly for $1331$ we have $1b^3+3b^2+3b+1=(b+1)^3$.
On
If $d_nd_{n-1}\ldots d_0$ is a string of digits, it is convenient to consider the polynomial $f(x)=d_nx^n+d_{n-1}x^{n-1}+\ldots+d_0$. Then $(d_nd_{n-1}\ldots d_0)_b=f(b)$. The OP's question is related to the question "Which integer polynomials take only composite values?".
The answers given so far describe two different phenomena (both illustrated in Mark Bennet's answer).
The Bunyakovsky conjecture would imply that these are the only reasons a polynomial takes only composite values.
Conjecture: Let $f(x)\in\mathbb{Z}[x]$ be non-constant with positive leading coefficient. Suppose:
$f(x)$ is irreducible, and
there is no integer $d>1$ such that $d\,\big|\,f(b)$ for all integers $b$.
Then $f(b)$ is prime for infinitely many positive integers $b$.
There are deterministic algorithms for checking whether a given polynomial satisfies the hypotheses of this conjecture.
$121$. Suppose it is prime in base $b$. Then $b^2+2b+1=(b+1)(b+1)$ is prime.