Unless you have it in writing from the MCU manufacturer that currents up to a specified limit may flow through the internal protection diodes during normal operation without adverse effects to other device parameters, on a modern MCU, its unwise to rely on the internal protection diodes for normal operation of your circuit. e.g. see
Microchip TB3013 for some of the issues that can occur. This is *NOT* hypothetical: over on the Microchip forums, there have been several reports of internal oscillator frequency shifts of 40% or more with less than 1mA flowing through a protection diode.
Long time constant capacitive filtering and an input that doesn't have a Schmitt trigger characteristic has its own problems. Normally they aren't electrical, though in low power designs the increased current consumption of the input gates due to an input voltage that spends too long in the transition region may be an issue, and I would be very suspicious of their possible behaviour at the top end of the MCU's permitted temperature range. However your code *MUST* cope with multiple random transitions due to noise at every actual signal transition. It would be a really *BAD* idea to use the input for an edge triggered interrupt, unless the ISR deferred re-enabling the interrupt for a time greater than the RC time constant. Its well worth reading Jack Ganssle's mini-series on
debouncing.
You also have to consider the sensing thresholds on the 12V side of the input protection network. A typical CMOS input will have logic thresholds near 30% and 70% of Vcc, with the actual transition somewhere between them. If you scale a 12V input down to logic levels with a potential divider, the percentage thresholds scale as well, so, assuming the divider ratio is chosen for 14V in, Vcc out, to best match the MCU logic levels to the 12V signal levels in a running vehicle, anything below 4.2V will be seen as logic '0', and anything above 9.8V will be seen as logic '1'. This applies as long as your input protection network doesn't rely on clamping to reduce the signal to logic levels.
If you use the simple opto input circuit: Opto-LED + series resistor on the 12V side and Opto-phototransistor to ground and pullup resistor to Vcc on the logic level side, the thresholds are a *LOT* harder to predict. They are determined mainly by the CTRR of the optocoupler, which is not a tightly controlled parameter and varies widely from device to device and with temperature and ageing. One approach to fixing this is to add a Zener in series with the LED to 'hold off' any significant current through it below the Zener voltage to guarantee the worst case transition wont be at an inconveniently low voltage. Another less accurate method is to feed the LED with a potential divider so that the voltage across it is below Vf for inputs below 4V or so.
The problem of transients has already been touched on - most opto LEDs have a peak pulsed If of typically an order of magnitude greater than their continuous rating, and if one follows the typical design practice of not exceeding 50% of the continuous rating in normal operation, that will permit the optocoupler to survive brief positive going transients up to around 250V if its setup for 14V input. However the permissible reverse bias on the LED is typically quite small, and although catastrophic avalanche breakdown is fairly unlikely, repeated negative going transients can degrade the LED's efficiency, reducing the CTRR. Its therefore advisable to protect it with an anti-parallel diode.
Protection against Load Dump transients is a whole different matter - worst case, your input circuit (and device PSU) can have to cope with a transient lasting half a second or so that peaks at over 100V. Power dissipation in your protection network during the surge is a significant issue. Fortunately modern vehicles tend to have load dump clamping at source to protect their ECU and other electronic systems, so unless your device is particularly high value, it may suffice to design it so it wont fail before the ECU does, and accept that you may get a few warranty claims from unlucky classic vehicle enthusiasts!