There are many ways to work with or on skills. Many skills such as research and some parts of engineering lend themselves to globalization. We might try the same for medical services and schooling. There already is medical tourism, and we might find it efficient to send kids to boarding schools in India. For parenting and farm work we do the reverse, we bring in cheap labour.
Skills and skill development that remains local could be self-regulated, controlled by supply and demand, or controlled by risk management such as through insurance and the courts (civil).
On the other hand, if we want skill applications (skill use) and skill development to be managed through skill engineering then we have to consider what tools we need. We can identify the need for measurement and for prediction.
Unfortunately we cannot measure skills but only skilled behaviour. Even a brain-scan does not show the wiring for the skill but only the brain-in-use while producing a skilled behaviour. Typically we measure with a number such as a grade in school. For a millenium or two that was the only way other than a description using language. Since the advent of computers we have more options. We can simulate the behaviour with artificial intelligence or even with robots.
We can see parts of education as a skills-acquisition project. We would like to measure the base and assess what tasks need to be done to move to the next level that can be verified by measurement. Given the base we make a prediction what the effect of a given treatment is likely to be. The prediction produces a hypothetical measurement. By comparing alternatives we can select the optimum treatment to apply. By measuring after applying the treatment we can assess the accuracy of the prediction.
Instead of taking a measure before and after, we can take a measure at regular intervals. For instance, instead of measuring at the beginning and at the end of a school year we could measure every day. Process control in chemical plants and manufacturing works by measuring much more frequently. These measurements can be used to detect undesirable deviations and to apply corrections immediately.
With a theory in the form of mathematical equations or computer programs we can make the prediction dependent on the context. The equations or programs allow us to represent the skill rather than the skilled behaviour, if we can find an appropriate measurement for the input. Validity of the representation now depend on correct preditions for all the situations in which the skill might be used.