i heard you like registers

approach

each instruction is split by a regex into a register, an inc/dec that becomes a small lambda, and the condition text after if. the condition is already valid python, so it's handed to eval with a defaultdict(int) of registers as the local namespace, which makes unknown registers read as 0.

the run keeps the register dict and the highest value ever written. part 1 is the largest register at the end; part 2 is that running maximum.

the trick

the conditions are python expressions as written, so eval against the registers does all the parsing.

def execute(instructions):
    context = defaultdict(int)
    mx = 0
    for var, op, cond in instructions:
        if eval(cond, None, context):
            context[var] = op(context[var])
            mx = max(mx, context[var])
    return context, mx