• Lockpicking: “Okay, solve this puzzle cube”
• Athletics: “Okay, lift this dumbbell”
• Attack: “Knife fight me irl, you coward!”

see i couldn't do this, if i were playing a character with high social skills, it's because i want the power fantasy ot being able to talk to people normally and convince them of things

if you ask me to do this IRL, i will probably end up four parentheticals deep explaining a tangentially related thing and talking about standard deviations or something

I mean, obviously I can't read the DM's mind, so maybe they are approaching it differently, but in my experience when the DM is asking "what do you say?" they're not looking to map your words 1:1 with what your character might. They want to get a sense of what sort of tone you're aiming for. What information you're emphasizing or omitting. If you might let some Important Tidbit slip accidentally that they can use as a hook. Keeping that in mind will probably help loosen your tongue.

One thing that has helped me immeasurably is knowing what my character sounds like. Their voice and cadence. Having that in your head (even if it's just a vague sense of it) is a massive help, and literally all my favourite characters to play, in D&D and OCs, have very clear voices. And it's not like I'm being original! Harlan, my neurotic lemur? Literally just Mordin Solus.

The last thing I'll touch on is that it's important to not assume "charisma = words". Charisma is also about presence and influence. It's getting people to do what you want. A massive warrior who speaks only in curt nods, but still manages to get people to do as they say because they're just so obviously powerful, is just as charismatic as a bard with a silver tongue. You can do an Oscar-winning job acting out a character, finding every exact word they'd say, but they might be a little turd that everyone hates and/or ignores.

So TL;DR, build up an archetype of your own character in your head, and worry more about closely hewing to that than what words you're saying (unless your DM's an ass), and you'll have a much better time with charismatic characters.

Neither of these options is always wrong. Consider: a game where you play a dexterity based minigame for dextrous tasks, a strength game for strength, endurance game for Con, etc? That would actually be pretty fun, even if not the thing for everyone. But that game obviously, to me, would not be the dragon game.

D&D is good for one thing: a particular kind of combat and puzzle navigating, resource management, and clever applications of class powers. What is horrible at and has never modeled in more than the most basic and almost useless way is persuasion, or other social interaction. It leaves that for players and GMs to work out on their own mostly.

There are better systems for it. But in my opinion, the more important thing to recognize is that there are other systems at all, and the scariness of "learning more rules" is also a D&D culture thing. Making learning un-scary is a task for the group to approach together. It's a skill that resides in the relationship between players in the group. In my opinion, these kinds of "what should be simulated, what should be decided, and what should players have to figure out and declare?" questions are actually best asked of the game system itself. They also need to be social agreements in the play group, and accommodate shifting needs. A player who usually wants to talk in character and have some back and forth that way before rolling anything social might have a day where they don't have that many words, but still want to play - and so on.

Consider how your table will handle situations of conflict between player and game. This is essentially a problem of consent. Like any play, it is only good if there is enthusiastic consent from all involved. Many rules in ttrpg play are a very kind form of CNC, consensual non-consent. We, as our character, didn't want this to happen, but we, as players, agreed this is the rule set and so the consequences of actions occur according to the rules. This, in principle, is why safe signals like the X Card are useful hacks. But regardless of specific signals, the conscious knowledge to be looking out for signs of the status of consent and degree of enthusiasm is the most important part.

If I want to say what my character says, and if the group wants me to, and the GM asks me to, there is not only no problem, things are perfect. If anyone isn't enthusiastically consenting and desiring this arrangement, something needs adjusting. This isn't something d&d rules teach you, and even many other games don't get into it even if they're reacting against that problem in d&d. Shout out to Flying Circus for having an excellent discussion of this in the core book. Copper "good effort" star to Lancer for having sections that come close to talking about this.

Seeking consent and desire is complex, but it's what makes fantasy play powerful and fulfilling. It's also the root of this "why are we simulating this that way but not that this way?" issue. It's not that one simulation is better always. It's that what gets simulated at all is a social agreement.

Building on the above: "Diplomacy" (or "Bluff," etc) being a flat DC to beat on a d20 + modifier is indeed a D&D-ass problem. Or a problem with so called "trad" TTRPG design in general.

There's so many PbtA and FitD examples of good social conflict mechanics it's not really worth listing the ones I know here. There's probably many more in other systems that I'm totally ignorant of. Social conflict can also be pure RP, which is honestly not my favorite but it's preferable to modeling social problems with the same mechanic as pit traps and sword-swinging, IMO.