... Product Hunt! We just shipped the biggest update to the Icons8 MCP since launch . The headline: illustrations . Your agent could already pick icons from Icons8, now it illustrates things too: a hero image for a landing page, empty states, onboarding, a 404. There are 345 styles , and the skill

... mistakes are unacceptable or irreversible. I think of it more like an intern: helpful with recurring work, but still operating within clear guardrails. Comment from Kuanze Ma(@kuanzema): @akarsh_hegde  In my experience, the secret sauce is creating agentskills that persist context and decisions in a Git repository so they can be reused across harnesses. That makes continuity portable and narrows the gap between harnesses in how effectively they manage context. Comment from Akarsh Hegde(@akarsh_hegde): Claude ...

... reporting. Better observability OpenTelemetry tracing, execution timelines, token usage, and model cost visibility. Easier workflow migration AI Convert can bring existing automations, including n8n workflows, into Heym without rebuilding everything manually. Built-in operational tooling Dashboards, DataTables, runbooks, and agentskills that can run Python and generate real files and reports. New integrations Amazon S3, Supabase, Discord, ElevenLabs, and ChatGPT through MCP. We are building Heym as a self-hosted AI workflow platform that teams can inspect, secure, and operate ...

... wasn't enough for the true scale people were aiming for. The "aha" moment actually came from looking inward at our own marketing team. We were already automating our own daily work using tools like Make, n8n, and various agentskills. We realized that the real power doesn't live in a design editor - it lives in your sheets, your files, and your data. So we thought: Why not just open up our core engine ... ...

... full control. Just bring your model and start talking to your data with Metabot, or building AI-powered workflows. That's not all, to celebrate, we also released several new AI goodies: Official MCP server Metabot in Slack AI AgentSkills And a lot more: https://www.metabase.com/release... Looking forward to seeing what you build ...

... OpenAI announced GPT-5.4 @AssemblyAI released Universal-3 Pro Streaming, their most accurate real-time STT model for voice agents @Superset is an IDE that allows you to run multiple coding agents

... community. Yesterday we launched ClawSecure and landed #2 Product of the Day ClawSecure is the free security scanner for OpenClaw AI agentskills. But I'm not here to pitch. I just want to share real traffic numbers and what I actually learned from our Product Hunt launch, so it's useful for other makers planning theirs 24-hour Product Hunt launch results #2 Product of the Day (Google Workspace CLI finished #3) 68 comments (most of any product launched ... ... rewriting our tagline to "antivirus for AI

... built SkillForge - a tool that turns screen recordings into reusable AI agentskills. **The Problem:** Creating browser automation scripts is tedious. You inspect elements, write selectors, handle edge cases, and maintain code as websites change. **The Solution:** Record your screen performing any workflow, and AI extracts the steps into a structured SKILL.md file that any AI

... everyone! Just shipped Build Skill, a CLI that scaffolds AI agent

... built SkillForge (skillforge.expert) to solve a problem I kept hitting: teaching AI agents my actual workflows. **How it works:** 1. Record your screen doing any task 2. AI analyzes and extracts reusable agent

... couldn't easily capture and reuse my workflows. Writing prompts is art, not science - you never know if you're giving the AI the right context. **The solution:** Record your screen doing any task, and let AI extract reusable agentskills from those recordings. It captures tacit knowledge - the unconscious decisions and patterns that make you effective. **What it does:** - Browser automation without coding - Reusable coding patterns - Data entry workflows - Teaching AI your personal preferences **Free to try:** 50 credits ...

... traces, rendered images, profiler output) is a first-class API concept the LLM proposer reads to make targeted fixes, and (2) Pareto-efficient search across metrics preserves specialized strengths instead of averaging them away. Results across 8 domains: learned agentskills pushing Claude Code to near-perfect accuracy simultaneously making it 47% faster, cloud scheduling algorithms cutting costs 40%, an evolved ARC-AGI

... reply to messages without touching my keyboard once. 2. App-Specific Profiles It context-switches for me. Outlook: Automatically polished to "Professional". Telegram: Switches to "Casual & Lowercase". VS Code: Formats code blocks automatically. 3. Ghost Morph (Agent

... give agents surprisingly broad access without losing sleep. Comment from Umair(@umairnadeem): i think most people in this thread are conflating "trust" with "giving full access" and those are completely different things. you dont need to trust an agent