if your app looks like the left one, it's a skill problem both same prompt & one shotted by astra right used appllama mcp + skills

MCP: Skills: Inspiration: Chani, Blank Spaces Prompt: /goal Use appllama mcp & skills, build an astrology app Stelari, it should have the flows of Chani and design it like Blank Spaces.

how many credits it burned on Astra and which? Ceedit Anxiety is becoming very real to me… sometimes you spend 45% 55% on the project in credit and and you get very bad results and sometimes I just use 5.6 Terra/High or Sol/High and get waaaay better results … I think it’s very weird

im on the 20x plan and it used like 7% of weekly usage

They’re simply not using applamma

it’s so easy 🤷‍♂️ tell em

Right app looks very identical to Endel's UI

now that you say, it does Lol blank spaces & endel twinnin

Looking amazing 🔥

Skill issue used to hurt my feelings. Now it's just a missing dependency.

Which image generation model have you used for the demo?

codex native imagegen, so gpt images 2

Just dropped this in a codex thread for an app I'm trying to get to testflight this week. I'll let you know how it goes

LESGOO 🦙 feel free to drop a dm anytime, we can work it together :)

right is a clear winner

Anything for web apps??

soon, in about ~4 months

Damn that looks really cool

lfggg 🙌🙌

I really love your product but $16 is a bit too much for me

Looks cool

I’ll give it a try, but seems too good to be true

sounds good, let me know how it goes. my dm is always open, i can guide through your building process :)

@vitorbukovitz0 Using it with my own prompts (agent-made), the AppLlama MCP + skills in Codex Astra Middle ngl improved my app a lot, but still unable to reach the lvl of your X posts. Would love if you can help me out a bit. Btw, banger product and fair, if not too little, for the value of it.

@vitorbukovitz0 sure, let's work it together :)

same prompt different skills is the clearest MCP demo I've seen today

a single prompt can produce vastly different results based on execution, fr

Share the repo or I don’t believe it

Skills are the interface between intent and execution

skills are the whole delta. same astra one-shot and the left still looks unfinished

Curious how much of the jump is the skill itself versus the system prompt behind it. Does the skill mainly enforce better defaults, or is it doing heavier lifting on structure and tone?

what specific skills made the difference for the right version?

The gap is usually in how the agent decomposes the task and what context it gets. MCP tools help, but the real win is defining the skill boundaries so the model doesn't have to guess. Curious what your skill definitions look like—do you keep them granular or bundle multiple steps

watched polished one shots collapse by iteration three; appllama needs editable structure, not prettier defaults

the scaffolding the world used to dismiss as pseudo research starts to clearly be something that boosts real world performance

an app's quality should reflect the skills behind it, not just tools used

right one hitting differently

is their a web version as well?

you must have a applama sub to use the mcp ?

Was the gap mostly from applama skills constraining layout, or from better component primitives in the MCP?

design skills are constraints. you limit yourself using them. giving picture references is the best thing you can do for your project design.

This is a great example of how much better the output gets with the right skills and MCP setup.

appllama mcp really makes the difference, those skill tweaks show up fast

Why is your product better than mobbin?

This makes the whole “anyone can build now” argument a lot more nuanced.

the before-and-after UI makes the skill gap obvious

I think both are not great, but the right one is def much better