As I see it, in AD&D there are three types of skills. (Of course, this analysis is my own inference and interpolation; the design principles are never stated in the books.)
1. There are some "highly reliable" skills that are successful in the range of 80-100%, such as the druid's nature lore, assassin's disguise, barbarian's first aid, ranger's tracking, and thief's climbing. Typically these skills fall into what we would call the exploration pillar today. They are reliable either because the consequences of failure are not very important (e.g. tracking, first aid) or so dire (e.g. disguise) that any large chance of failure would make them not worth using.
2. Then there are some "difficult, can't even try if untrained skills" of which the thief skills are the archetype, but there are a few other examples such as bard charm & lore, and barbarian detection of magic and illusion. The design principle for these skills seems to be that succeeding gives you an advantage, but failing doesn't have a major consequence. A thief who fails detect traps is no worse off than before--they may still get a 1-2 in 6 roll to avoid (a frequent mechanic in modules) or use a creative approach to mitigate the risk. Failing to hide/move silently does not mean you are detected, it just means your chance of surprise remains the default. At least, that is how I ran it--if failing a detect traps roll meant setting off the trap, no one would ever try it!
3. The middle ground is when PCs attempt challenging tasks that require no special training. In 1e, we have some d6-based examples of such tasks, like open doors and the implicit perception vs stealth opposed check of surprise. ....
