Script to mirror transform similar to Maya's Joint Mirror Tool.

xformMirror.py

import pymel.core as pm

def xformMirror(transforms=[], across='YZ', behaviour=True):
    """ Mirrors transform across hyperplane. 
    
    transforms -- list of Transform or string.
    across -- plane which to mirror across.
    behaviour -- bool 
    
    """  
    
    # No specified transforms, so will get selection
    if not transforms:
        transforms = pm.selected(type='transform')
        
    # Check to see all provided objects is an instance of pymel transform node,
    elif not all(map(lambda x: isinstance(x, pm.nt.Transform), transforms)):
        raise ValueError("Passed node which wasn't of type: Transform")
    
    # Validate plane which to mirror across,
    if not across in ('XY', 'YZ', 'XZ'): 
        raise ValueError("Keyword Argument: 'across' not of accepted value ('XY', 'YZ', 'XZ').")        
    
    for transform in transforms:
    
        # Get the worldspace matrix, as a list of 16 float values
        mtx = pm.xform(transform, q=True, ws=True, m=True)
    
        # Invert rotation columns,
        rx = [n * -1 for n in mtx[0:9:4]]
        ry = [n * -1 for n in mtx[1:10:4]]
        rz = [n * -1 for n in mtx[2:11:4]]
        
        # Invert translation row,
        t = [n * -1 for n in mtx[12:15]]
        
        # Set matrix based on given plane, and whether to include behaviour or not.
        if across is 'XY':
            mtx[14] = t[2]    # set inverse of the Z translation
            
            # Set inverse of all rotation columns but for the one we've set translate to.
            if behaviour:
                mtx[0:9:4] = rx
                mtx[1:10:4] = ry
                
        elif across is 'YZ':
            mtx[12] = t[0]    # set inverse of the X translation
            
            if behaviour:
                mtx[1:10:4] = ry
                mtx[2:11:4] = rz
        else:
            mtx[13] = t[1]    # set inverse of the Y translation
            
            if behaviour:
                mtx[0:9:4] = rx
                mtx[2:11:4] = rz
         
        # Finally set matrix for transform,       
        pm.xform(transform, ws=True, m=mtx)

This is great - thanks for sharing. The only suggestion i would make is to declare a dictionary to store the transforms, then apply the transforms in a second pass... so replacing line 60 with this....

    stored_matrices[transform] = mtx

for transform in transforms:
    pm.xform(transform, ws=True, m=stored_matrices[transform])

As the only issue i have found is selecting a hierarchy of joints, its reading the transforms whilst its setting them which can cause issues - which caching the transforms first means you're always reading the right worldspace transform