Array Escape - get out of there

Python, 161 138 bytes

Credits for factorial.

g=lambda x:0**x or x*g(x-1)
f=lambda a,i,n=0,l=[]:(i<0)+(i>=len(a))and n or(0 if i in l else f(a,[a[i],i-a[i]][i and-g(i-1)%i],n+1,l+[i]))

Ideone it!

How it works

Wilson's theorem is used for prime checking.

Loop detection by storing seen indices to an array (l) and checking whether current index is in l.

Python, 107 bytes

import sympy
f=lambda a,i,n=0:0if n>len(a)else f(a,[a[i],i-a[i]][sympy.isprime(i)],n+1)if 0<=i<len(a)else n

Usage: f(list, start) ex: f([2,5,6,8,1,2,3], 3)

Returns 0 for loops (detected when n > len(a))

Matlab, 138 bytes

This a straighforward approach, using 1-based indices because Matlab uses 1-based indices by default. To count the number of steps we use a for loop counting from 1 to infinity(!). For the case were we cannot escape the array, we use a vector v to keep track of which entries we already visited. If we visit an entry twice, we know we are stuck in an unescapeable cycle. To see check whether we are outside of an array, we use the try/catch structure, which also catches out of bounds exceptions.

function r=f(a,i);v=a*0;v(i)=1;for k=1:Inf;if isprime(i);i=i-a(i);else;i=a(i);end;try;if v(i);r=0;break;end;v(i)=1;catch;r=k;break;end;end