Jarmo Tuisk

I am a little bit scared of this. Have you checked potential attack vectors? I am not a cyber security expert, but someone injecting a takeover prompt and then moving easily around in my compter and browser, seems terrifying. I am building native app and I would benefit hugely from automated testing though.

Oh that's the nice one. I'm building native apps for multiple platforms (Ritemark for Mac Arm, x64, Windows) and build times are loooong. CI failures sometimes happen like 1h into build. Will auto-fix help me in those scenarios?

Smart approach. The gap between "I see a component I like" and getting my agent to reproduce it is still way too big. Screenshots lose all the interactive behavior. Does it capture state and interactions too, or mainly the visual structure?

Ritemark - Office for Your Agents

Can you bring your own context or instructions to the agent, or does it start fresh each time? For anything beyond a landing page the agent needs project context to make good decisions, and that's where most of these tools break down. Building Ritemark app - context and instructions are piling up as fast as code.

The "without the MCP context tax" line caught my eye — that's a real pain point. We have 100+ skills/tools defined across our agent setup too and most of the work isn't adding new ones, it's keeping the existing ones from going stale or contradicting each other. Curious if anyone else is hitting "skill debt" yet — like tech debt but for agent tooling. You add skills fast, they rot quietly, and...

this is a smart approach — the MCP integration means any AI that speaks MCP can publish directly, which removes the biggest friction point in the "vibe coding" workflow: actually getting your stuff live. we build with Claude Code and the amount of throwaway HTML that gets generated in conversations but never sees the light of day is wild. having a one-step publish would save a lot of...

oh man this hits close. been building a desktop app with Claude Code for months and the CLAUDE.md staleness problem is probably our #1 source of wasted time. you write great instructions, they work for a week, then the codebase evolves and the agent starts doing weird things because the context is outdated. we ended up with this whole system of CLAUDE.md + AGENTS.md + skills files and honestly...

been building with Claude Code for months now and the "quick skim" problem is very real. agents write code fast but the subtle bugs pile up — especially when one agent changes something another agent built two weeks ago. multi-agent review makes a lot of sense here, curious how it handles context across larger PRs where the full picture only emerges from reading multiple files together.

I do AI trainings for teams so this trust thing comes up like every single session. Agents are getting really good but honestly how much you trust them depends way more on how you set up the context — guardrails, instructions, steering — than on the model itself. Like maybe 30% is the model and 70% is your prep work. Best analogy I have is hiring a summer intern from college. Smart kid, learns...

This hits close to home — I'm building a markdown editor (Ritemark) and one of the biggest decisions right now is exactly this: do we integrate Claude SDK directly, or go through something like 21st Agents SDK? Direct integration gives us full control but we're spending a lot of time on the plumbing — streaming, session handling, error states. Stuff that has nothing to do with what makes our...