My advice: If you are going to run a skill challenge you need to flesh it out. A lot. So that you are not trying to come up with descriptions of things off the top of your head and making stuff up on the fly. Some people might be able to pull that off. But I think you have to be prepared for improvisation. And that is coming from someone who has been involved in improvisational theatre for the last 6 years of my life.
Which equates to putting in some work before the session.
I do it this way at the most basic level:
My skill challenge will have 3 phases. I usually begin with an initial set up phase. This can go on before the PCs are aware they are even in a challenge. An example from my current pbp game: The PCs are headed towards their destination the horizon where there is a storm brewing (first clue), two junks sail lazily towards the boat (second clue). The PCs goal in this phase is to discover the imminent attack from the pirates on the junks before it occurs. If they do so they will achieve 2 successes. I don't care how many rolls they make within that period, the goal is the discovery of what is going on. If they don't discover the attack in time they will have 1 failure.
How to proceed from here? My players have to tell me what it is they are doing. If they pay attention to the clues then they have a chance of success. If they don't ... well then things move forward by my hand, they find themselves at a disadvantage tactically and they have 1 failure for the challenge.
If the players ask me about the junks then I ask them to make a roll. Here I can give them a choice (and you would have this part prepared before hand) of possible relevant checks. Eg. Streetwise, Perception, Insight
My DCs will be mostly medium DCs but additional information (and more important and revealing information) will be revealed with higher DCs.
My streetwise medium DC reveals that the boat has the symbol of traders of precious stones from the Jade Isles. The hard DC reveals that the symbol is actually reversed. Reversing a symbol in these parts is a way that the Yakuza mark which ships are not to be robbed. The information in itself still requires the players to come to their own conclusions.
But when contrasted with the information from Insight (the boats definitely changed tack and headed directly towards you when they spotted you) or perception medium (a figure on one of the boats stands on the prow waving to your ship) Hard (you catch the glint of steel on the boats; the sailors on those boats are very well armed) then the players begin to smell rotten fish. And more so if the take a look at the storm on the horizon which is actually a naval battle.
So they get a few rounds of actions to check things out as the boats continue to close the distance. I know in my head that figuring it out before the attack is worth 1 success, and that alerting the captain and crew in time is worth the 2nd success. The rolls, whether successful or not are largely inconsequential. It's the achievement of the goals that is the challenge.
My players discover the ruse, alert the captain and then the junks make their move as the change their sails and drop low into the water preparing for their attack. At which point I announce that they are in a challenge and that they currently have 2 successes and no fails. The next phase presents a choice for the PCs fight or flight. This is the Development phase of the initial situation. Each is a completely viable choice to complete the challenge successfully, each has their own mechanics which I have prepared before hand. Each is a type of mini challenge requiring 3 success before 3 fails. Escaping involves outracing the 2 ships in a race to the nearby port. PCs can assist the Nature roll of the navigator in ways that make nautical sense. They also have the opportunity to man the cannons and take pot shots at the boats and try and leave them lame in the water. Or they can choose to fight with options to either board the ship and subdue the captain and the first mate forcing the crew to surrender or outmanoevre the ship and ram into it's side. A high enough roll on the ram could see the PCs boat pass straight through the enemy ship. Now this was a for a pbp game so the format is different, but for a table top game I would do this by giving the players a hand out of their options depending on whether they had chosen fight or flight.
Now here again I don't care how many rolls they make or whether they are successful or failed rolls. 1 success for either outrunning each boat or crippling each boat, subduing each boat or a combination of the two. Again for a total of 2 more possible success that can be achieved during a phase.
Now failure on the other hand is marked by the state of their boat. Their boat can take 8 hits. When it has taken 4 hits (bloodied) the have 1 fail. If their boat takes 8 hits it is crippled and they get boarded for a second fail. Also when the boat is hit by cannon fire then I make a secondary attack roll against everyone on board who takes splinter damage.
Then I add social elements so that the players can order around the crew or have them help man cannons with minor actions using charisma based skills. I continue to develop the situation as the junks fire out massive grappling hooks which slow their boat and either need to be cut away (3 hits) or unhooked with thievery. The junks have a shugenja aboard who concentrates their magic on the boat navigator dazing and eventually stunning him into inaction so one of the PCs has to take the helm. Here arcane PCs can use their knowledge to protect the captains mind with minor action arcane checks. Once hooked the boats will have contests using nature skills. If the junks win the contest the pull closer and closer as they near to board. Meanwhile cannons are fired on both sides and ships are weakend.
So my PCs overcame this part with 1 hit left remaining on their boat, several PCs bloodied. They chose to fight and with a combination of cannon fire, improvisation with PC ranged thunder powers they left it crippled. The second boat got its grappling hooks into them and began to pull closer but they blasted it enough to cripple it as well before they were crippled themselves.
And so I began phase 3. This is where I put my Twist of Events. In this case they spot that the enemy boat has prisoners aboard tied to the mast. they wear the colours of their benefactor. Now they can choose, flight or rescue. the rescue will be how they achieve the next two successes. There are 3 prisoners aboard. One appears dead. They can get 1 success per prisoner rescued. It is a complexity 2 challenge so they need 6 successes. They already have 4. They can achieve a fail for each prisoner killed. Or they could abandon and simply gain a partial success for the challenge which is basically success in so far as moving the adventure along but a failure on certain minor quests that can only be completed by saving the prisoners.
The final part involved them boarding the pirate junk and making their way through the swarm pirate crew who set upon them and challenged movement. They could still subdue the crew by bloodying the captain and the first mate. It got pretty grisly but they pulled it off and even managed to complete a bonus minor quest which was to rescue the third prisoner trapped beneath the fallen mast.
Again, to reiterate, the rolling of dice and attempting things was only important as to approaching their goal. It wasn't: I roll athletics ... a 2 ... 1 fail ... aww crap. Or ... I don't want to roll stealth (in your case) because I don't have training we'll get a fail if I do.
It is important to note, especially for a table top game, that you never let your players get away with "I use Diplomacy". I lay it out this way to my players. You don't get to roll anything until I ask you to roll. You tell me what it is your PC is doing and I will decide whether a roll is necessary and depending on what you have described your PC doing I will choose what roll you need to make. I will often lower DCs to easy DCs for really cool ideas or good rolplay or just grant an auto success.
Also if the goal is only to roll high you will only ever wind up with players looking for their highest skill and finding ways to convince you that doing a backwards flip will somehow help you in your negotiations with the prince. Goals should mark the points of success in the challenge. So you need to have clear, pre-prepared, visualised, annotated or sketched out before the session begins.
If you are interested in seeing how the skill challenge described above played out here is the link:
http://www.enworld.org/forum/living...l-lovers-dm-jbear-judge-lord-sessadore-8.html The build up begins there. The challenge is announced 2 pages later.
The challenge is making it work in a table top game. This can be done with carefully prepared and easy to understand and read handouts for each of the players. Again .. its an investment in time (and printing ink), though not all SChallenges have to be such a complex mix of combat and abstract as the one I use as an example.
The key element for me is the phases, clear goals which are the marks for both success and failure not individual rolls, development of the situation that builds tension, throw a spanner in the works at some stage, have consequences that again alter the situation and add to the tension and pressure if the PCs gain a fail. That is what is at the heart of making skill challenges work for me.