Does this skill help?
It runs a task twice. Once with the skill, once without it. Then it compares the two runs.
When the difference is too small to call, or the sample too thin, it says so. It does not fill the gap with a number.
One skill, measured
Each run of the task is one try. The same task runs eight times with the skill and eight times without, and the two sets are compared. The task hides a false fact on purpose, so a try either catches it or misses it.
Two tries apart out of eight. That is not enough to tell the two sets apart, so no verdict is recorded. This is a result, not a failure. Widen the sample and the instrument will call it.
Where to go next
Published determinations
Every skill measured so far, and the ones still waiting on evidence.
What the instrument found
All results to date, including the runs that came back undetermined.
Run it on your own skill
Install the package and audit a skill file on your own machine.