Koj, Daijs (neither has あとう). Koj has 適う as well.

That's certainly a solution. I was a little reluctant  as あとう is in Daijirin and GG5.

@@ -6,0 +7,3 @@
+<k_ele>
+<keb>適う</keb>
+</k_ele>
@@ -9,3 +11,0 @@
-</r_ele>
-<r_ele>
-<reb>あとう</reb>
@@ -17,0 +18 @@
+<s_inf>plain form pronounced あとう（アトー）</s_inf>

remove あとう from reading and add [note="dictionary form pronounced あとう（アトー）"]?

I think the options here are:
(a) do nothing. (It's a very marginal situation)
(b) split 能う/あとう off into a v5u-s entry as was done with 給う. (I'm not sure it really behaves like the other v5u-s verbs.)
(c) create a special reading tag ("no-infl"?)
(d) add a note, e.g. "あとう does not inflect".
My inclination is to do (d), although (a) is tempting. Any other suggestions?

koj

@@ -14,0 +15 @@
+<pos>&vi;</pos>

meikyo: 表記終止形・連体形では「アトー」と発音し、「あとう」と書くことが多い。

I think we should have a label for such ウ音便 readings as あとう, to indicate that it is to be pronounced アトー and not to be conjugated further.
(Only a few verbs had their ウ音便 forms split into a new stem+ending combination, e.g. 酔う（えう→よう） and 言う（いう→ゆう）. Most are pronounced as a single morpheme, despite Daijirin's cross-reference entry formatting.)

It is simpler to label the ウ音便 readings than splitting them to a new entry, as we did with 給う. Doing so also avoids duplication and overloading the [v5u-s] label.