I've seen the example of sequential workspaces. I was wondering if you have an ouput that you've reviewed is there a way to trigger the subsequent step without having to type in "now do step 3"? I am trying to develop in cowork because my team doesn't have claude code access - and IT is being careful about handing out those licenses.

💰 $325 CASH 💰 That's a full year of Premium. Win this and your membership pays for itself. But the real prize this week isn't the cash. Keep reading. 📋 THE CHALLENGE You just got hired again. Meet Diana, owner of a 4-person boutique real estate team in Austin. 60-80 transactions a year, mostly residential, mix of buyers and sellers. 📎 Download the full client brief attached to this post. Short version: She doesn't want software. She wants a system she can teach her team to use in a week. Your job is to build the AI operating system for her team. This isn't one specialist. This is a small team of AI specialists organized into a multi-folder ICM architecture, with a clear handoff protocol between them. 🗂️ WHAT YOU'RE BUILDING Last week was one specialist. This week is a team of them. Required folders: 📍 00_orchestrator/ — The router. Where every request starts. Decides which specialist gets the job. 📍 01_lead_qualifier/ — First contact with new prospects. Captures intent, budget, timeline. 📍 02_property_research/ — Deep research on specific properties or neighborhoods. 📍 03_client_communication/ — Drafts emails, texts, follow-ups in the voice of the agent. 📍 04_transaction_coordinator/ — Handles the deal once it's live. Checklists, deadlines, document tracking. Each folder must include: - 📄 identity.md - 📐 rules.md - 💬 examples.md - 🔗 handoff.md (NEW for Week 4 — how does this folder pass work to another folder?) - Plus a root-level README.md explaining the architecture, the typical flow, and how to onboard a new team member. 🔥 WHY THIS ONE IS DIFFERENT Weeks 1, 2, and 3 were warmups. This is the comp where the work you ship genuinely starts to look like the real thing. The handoff protocol is the test. Anyone can build five folders. The hard part is defining what each one needs from the previous one and what it passes to the next one. That's where multi-agent systems actually live or die.

Before I get to the who and why, I want to say this plainly. Picking a winner this round was genuinely hard. I went through every submission. Pulled repos. Read identity files. Compared rules.md sections. Every entry did real work. A lot of you shipped something I would happily use, sell, or hand a client tomorrow. I can NOT explain to you how proud I am of everyone participating in these, you make this community worth and it and there is SO much potential for the future from just ONE competition let alone future ones. 💼 That part matters more than the prize. The $325 covers a year of Premium and that's great but... The portfolio piece is the real value here. A public repo of a working folder-based AI specialist with a clean README and receipts is a resume line that hiring managers can clone and test cold in five minutes. It is also something you can charge for. If you built a specialist that solves a real problem in a real domain, you already have most of what you need to license it to a peer in your industry, sell it as a service, or package it as a Done-with-You engagement. A few quick notes on that, because most of you did not realize what you actually built and I want to Highlight a few of you: 💸 @Nicolas Patron Uriburu USD Routing Coach AR could be sold to every Argentine indie consultor I know. Same playbook works for any country with FX restrictions. Subscription service, recalibrated annually, audit-pack included. It is a product. 🔗 https://github.com/Nicopatron/usd-routing-coach-ar 🔧 @Jannetje van Leeuwen RAMS specialist is a service business in waiting. Irish signage contractors will pay for this. Same model works for any trade with a regulatory documentation burden. Plumbing, electrical, fit-out, demolition. Each one needs its own folder. 🔗 https://github.com/JannetjeIQ/rams-irish-signage-installer

The pitch I walked into the room with wasn't a deck. It was a working pipeline running on my own Instagram account, and a folder I could rename. Here's the actual thing. The workflow (built for @ari_leavesley.md first) Every carousel post on my personal account runs through the same folder structure: 1. _config/. Brand identity, voice rules, visual style, compliance. Loaded once, referenced everywhere. 2. CONTENT_LOG.md. Raw capture. Anything I notice, ship, break, or learn. No formatting pressure. 3. stages/00_mine/. A pass over the log that surfaces carousel candidates. Looks for the anomaly, not the achievement. 4. stages/01_draft/. Each candidate gets turned into a slide sequence. Hook, tension, payoff, receipt. Same skeleton every time. 5. stages/02_render/. Drafts hit a Remotion template. Dark background, monospace header, watermark, final slide. The visual system is one file. 6. stages/03_qc/. Compliance check. Hashtag pool, engagement bait, platform rules. Blocks anything that fails. 7. posted/. Archive. Feeds analytics back to _config when patterns emerge. None of that is novel on its own. What matters is that every stage reads from _config/ and writes to the next stage. One job per folder. Plain text between them. Why the port took an afternoon When the business asked "can you do this for us," I didn't rebuild anything. I copied the workspace folder, renamed it, and swapped three files: - _config/brand.md. Their identity, not mine. - _config/voice.md. Their tone, their vocabulary, their do-not-say list. - _config/visual.md. Their typography, their palette, their template. The mining stage, the drafting stage, the render stage, the QC stage. None of them changed. They don't care whose brand it is. They operate on whatever _config/ hands them. This is the part I keep trying to explain to people who think AI content is about finding the right prompt.

Tired of Claude opening every piece of feedback with how great your writing is before getting to what's actually wrong with it? Me too, so I built a skill to fix that. When triggered, it skips the praise and: - Opens with a revision score (0–100%) - based on how much work the piece needs - Leads with the three biggest structural problems (not typos) - Quotes specific lines and says what's wrong with each one - Ends with a full rewrite so you leave with something usable I didn't find anything in the skill repos that worked the way I wanted, so I made one. Dropping it here - use it, adapt it, credit appreciated if you repost. A few things I'm genuinely curious about: 1. If you try it - what would you change? 2. If you've handled the sycophancy problem a different way (custom instructions, system prompts, something else entirely) - what's actually working for you? 3. While there are a metric F-ton of skill repos out there, would there be a benefit to having a stored set of skills and such in Clief Notes? We could vote (up/down) on the skill/plug-in/etc. with the highest rated making it into the repo... critical-editor Skill ------------- Brutally honest, anti-sycophantic editorial review of written work. Use this skill whenever USER asks for feedback, editing, critique, review, or a second opinion on any written content - emails, proposals, reports, slides, marketing copy, strategy docs, or any prose. Also trigger when he says "red team this", "tear this apart", "be brutal", "don't hold back", or "what's wrong with this". This skill overrides Claude's default tendency to lead with praise and forces problems-first output. Use it even if the request sounds casual like "take a look at this" or "thoughts?" - if content is attached or pasted, apply this review protocol. Critical Editor Your job is to tell USER what is wrong with his work before anything else. The default instinct to open with encouragement or frame criticism softly actively harms the quality of feedback. A good editor earns trust by being accurate first, kind second.