Sine, cosine and tangent all fail to be one-to-one over their full domains, so each is first restricted to an interval on which it is a bijection, and the inverse is defined only on that restricted piece.
Inverse sine. Restrict sinx to [−2π,2π] (the interval on which it is one-to-one AND onto [−1,1]). Then sin−1:[−1,1]→[−2π,2π] is defined by sin−1x=y⟺siny=x and y∈[−2π,2π]. Its graph is the mirror image of the restricted sine curve in the line y=x: an increasing S-shaped curve through the origin, from (−1,−2π) to (1,2π). Since sine is odd on its restricted domain, sin−1x is also odd: sin−1(−x)=−sin−1x.
Inverse cosine. Restrict cosx to [0,π] (one-to-one, onto [−1,1]). Then cos−1:[−1,1]→[0,π] is defined by cos−1x=y⟺cosy=x and y∈[0,π]. Its graph falls from (−1,π) through (0,2π) to (1,0). Cosine on [0,π] is neither even nor odd (evenness would need a domain symmetric about 0 where it stays one-to-one, which is impossible for a periodic function away from the origin), so cos−1x is neither even nor odd either; a useful substitute identity is cos−1(−x)=π−cos−1x.
Inverse tangent. Restrict tanx to (−2π,2π) (one-to-one, onto all of R since tangent has vertical asymptotes exactly at the endpoints). Then tan−1:R→(−2π,2π) is defined by tan−1x=y⟺tany=x and y∈(−2π,2π). Its graph is a flattening S-curve through the origin that hugs the two horizontal asymptotes y=±2π without ever touching them, since tan−1x is defined for every real x (unlike sin−1,cos−1, whose domain stops at ±1). Tangent is odd on its restricted domain, so tan−1x is odd. …