Agility based evasion is a bit weird from a missile on missile evasion perspective because it assumes that missiles somehow have the ability to decide to engage in evasive maneuvers. Understand that because of the small detection range of AMMs combined with the relatively large attack range of attack missiles means that missiles will somehow need to evade missile which they don't even know they exist. So such detection would require an onboard missile sensor, passive or active which in turn takes up a large part of the missiles real estate, kind of defeating the point of the exercise.
Mind you this is a realism argument, not a gameplay one. I think from that perspective the missile needs to know that the AMM exists for it to even be able to attempt evasive maneuvers.
You could just say that the missiles flies erraticly to avoid being an easy target. And higher agility allows for better maneuvers/a better threat detection system while still staying on target.
This is exactly what "real" missiles do. If you look at the literature on designing control schemes for anti-missile missiles, the key parameters are the off-axis dynamics of the interceptor and target. In much of the literature, this is assumed to be a ground-launched interceptor and a nuclear reentry vehicle. The reentry vehicle will at some point (either randomly or at a carefully chosen moment depending on the reentry vehicle designer's knowledge of the interceptor's capabilities) perform a evasive maneuver (usually a max acceleration turn, but again the optimal solution depends on what you know about your opponent). The interceptor must then attempt to follow the maneuver; its ability to do so successfully depends on the ratio of off-axis accelerations achievable as well as time delays (primarily actuator lags, but can also include sensor issues).
Agility is thus an abstraction of missile tonnage devoted to off-axis thrusters, sophisticated computers necessary to solve the differential game in real time, and the g-hardening necessary to allow aggressive maneuvers without breaking the missile. And the thing to note is that the effects of this on the interception encounter are symmetric; making the reentry vehicle have a smaller time constant is simply the inverse of making the interceptor have a larger time constant. Thus, a decent model for how agility affects accuracy would be to take the ratio of interceptor and evader agility, rather than the current model of just using the interceptor's agility.
Agility should also affect beam weapon accuracy, but beam weapon accuracy would also need to account for target size for that to really make sense. Size doesn't matter so much for missile hitting things because the lethal radius is generally much larger than the target...it doesn't matter if the target is 2 meters or 3 meters when the missile is lethal if it blows up within 10 km of said target. But a beam weapon needs a direct hit to do anything at all, so a 2 meter cross section is meaningfully more difficult to hit than a 3m cross section would be.
Making smaller ships harder to hit might make beam fighters actually effective, as well as just generally giving a better reason to build small ships. There are economic advantages to smaller vessels, but in terms of performance per ton, bigger is better in Aurora (at least for beam combat; missile range scaling with resolution gives small to mid-size vessels a niche there).