If you take a closer look at the popular decks here on Arkhamdb, you may notice that still most of them use a complete set of Guts, Perception, Overpower and Manual Dexterity. I omit Unexpected Courage here, because it functions a bit different.

These cards' popularity is based on their marginal cost and the high versatility they give you. Each of them occupies one deck slot, but there is (on Standard difficulty) a huge chance of drawing another card, so it feels as if they cost nothing. They kind of "replace themselves instantly". Whatever strategy you are running, tests over all four attributes will meet you at every corner (or in every Mythos phase) and you don't have to commit these skills to a particular kind of skill test. The above-mentioned reasons, as a sum, virtually "reduce" your deck size and enable you to draw your key cards much faster. In fact, playing them includes two slight risks:

• The test can fail and the skill does not replace itself.
• The skill succeeeds and the card you draw is your Weakness. Especially, this risk can hurt if you Weakness happens to be Amnesia.

Pros

• +2 on a test for (almost) no tempo hit is really, really good. If taking the test costs you an action, and you're even on the difficulty, then the resulting tempo bump is, on average, worth almost an entire action.
• Can be committed to another player's skill test.
• Zero play cost
• Card text has no trigger restriction (as in Vicious Blow e.g.)
• Combines perfectly with cards that have an additional effect if you overfulfill: List
• "Ashcan" Pete gets more value out of the skill card than the other investigators, since the extra card you draw also works as smelling salts to wake his dog up.
• On High/Expert difficulty you do not want to fail on skill tests, so the +2 bonus alone justifies the inclusion.

Cons

• "Max 1 committed per skill test" across all players.
• The skill succeeeds and the card you draw is a Weakness.

Unless my math is wrong, Perception, Guts, Overpower, and Manual Dexterity are by far most cost efficient and overpowered cards in this game. Usually a good card is something that gives us an equivalent of getting 1.5 actions per 1 action (example: 1 action to draw Emergency Cache, 1 action to play it, and you get equivalent of drawing 3 resources). But with Perception, as an example we'll take The Gathering scenario standard difficulty:

• If test difficulty is one higher than our skill, Perception gives us +400% chance of success against 37.5% chance of not replacing Perception card (10.67 per 1 card efficiency).
• If test difficulty is same as our skill, Perception gives us +116.67% chance of success against 18.75% chance of not replacing Perception card (6.22 per 1 card efficiency).
• If test difficulty is one lower than our skill, Perception gives us +40% chance of success against 12.5% chance of not replacing Perception card (3.2 per 1 card efficiency).

How i calculated numbers above ^, example: assuming no ghouls in your location, taking one action to retry ability 4 against difficulty 3 test without committing resources would give us a chance of drawing 10 success chaos tokens from 16 total tokens, or 62.5% of success, with Perception committed it's 87.5%. The chance to lose Perception card is the amount of losing tokens (2) against total tokens (16) = 12.5%

And that's not even taking into account its deck thinning ability, Underworld Support for example costs 1 XP (equivalent of 0.5 actions), but 5-8 of these cards instead will give you the same deck thinning 1) without XP cost, 2) without limiting good cards to one copy, and 3) you can commit these to other players.

The upgraded versions of these cards are not nearly as cost efficient.