CLAUDE COMMUNITY CHICAGO
The workshop at 1871, rebuilt so you can make every piece yourself. Two builds, the deploy, the four agents as real files, and your credits.
I'm an introvert, and networking has always been a cap for me, so I built the follow-up I'd never remember to do on my own. Something captures a person I just met, Claude fills in the gaps and drafts a short specific note overnight, and I wake up, read it, and hit send. Full credit to Tina Lopez, who showed me a version of this first, because that's what the night was really about, you solve your own problem and someone else in the room has the same one.
In a claude.ai chat, and if you have an iPhone the capture step becomes an Apple Shortcut so you can do it from your pocket. No iPhone is fine too, since a notes app and this saved prompt do the same job, and the habit is the part that matters.
I want to build a simple system for following up with people I meet. How I meet people: [conferences, church, client meetings, wherever you actually meet them] How I want a follow-up to sound: [warm and short, or formal, or however you write] Where I keep contacts today: [phone contacts, a spreadsheet, a CRM, nowhere] Help me design a capture-and-follow-up system that: 1. Captures a name, where we met, and one thing they said, in under thirty seconds. 2. Uses that to fill in their role, company, and context, and tells me when it is guessing. 3. Drafts a short, specific follow-up message in my voice, using the one thing they said. 4. Leaves it for me to read and send myself. It never sends on its own. If I have an iPhone, walk me through building it as an Apple Shortcut step by step, and write the exact prompt the Shortcut sends you each time someone new gets captured. If I do not, give me the saved prompt I paste in each time, with the three blanks I fill.
The two lines worth editing are the voice line (make it sound like you, and paste in a real message you've sent if you want it exact) and step 3's "in my voice", which you can widen to "and suggest one specific next step" once the basic version is running.
Everything I do sits behind one link in my bio, and I built it with Claude for free. One page, your name, what you do, and the three or four places you want people to go.
Built in a claude.ai chat, downloaded as one HTML file, and put on the internet in section 3. No code, no framework, no monthly fee.
Build me a single-page HTML link-in-bio site. Who I am and what I do: [describe yourself in two or three sentences] What I want linked, in order: [Instagram, LinkedIn, my site, my calendar link, whatever you want people to click, most important first] The one thing I most want people to do: [book a call, follow me, buy the book] Colors and feel: [clean and dark, warm and bright, minimal, or name a site you like] Rules: - One HTML file with all the CSS inline. No frameworks, no external scripts. - Looks good on a phone first, since that is where people will open it. - The one thing I most want people to do is the biggest button. - Ask me anything you are unsure about before you start, then give me the finished file to download.
index.html. The filename matters for section 3.index.html on your own computer, it opens in your browser looking right, and every link goes where it should.Come back to the same chat any time and say "add a link to X" or "change the colors to Y", then download the file again. Section 3 is where you replace the old one.
That page Claude just built you is a single HTML file, so getting it live needs no terminal and no install, and it takes about three minutes on Cloudflare Pages, which is free.
In your browser, at dash.cloudflare.com. You'll need a free account, and the email confirmation is the slowest part.
index.html in a folder on its own, and name the folder anything you like.To update the page later, go back to the same project, click Create deployment, and drag the folder in again. To put your own domain on it, the Custom domains tab on that project does it from the same screen and you won't have to rebuild anything. No affiliate link anywhere in this, it's just genuinely free.
These are the four agents that did the most to make my life with AI better, and on stage all you got was their names. Here they are as files you can drop in and run today. A chief of staff that reads what changed and tells me what needs me, a skill builder that turns anything I've explained twice into a tool, a completeness auditor that catches what's missing before anything ships, and a stress tester whose whole job is to say no. Each one is a trimmed-down version of the file I actually run, with my private setup stripped out and the behavior left in.
Pick where you build. The choice sticks, and you can flip it any time.
You get two ways in for each agent. The fast one is pasting the instructions block into a Project, and the durable one is uploading the Skill file so Claude reaches for it on its own. Do the first tonight and the second when you've used it twice.
You get one .md file per agent in the exact shape Claude Code loads. Drop it in the .claude/agents/ folder of any project and ask for it by name.
Sits between everything getting built and what reaches you. Reads what changed, flags what you're not seeing, puts a confidence level on every status claim, and tells you what needs you today with one recommendation per item.
Loading the instructions from files/claude-ai/chief-of-staff-instructions.md
chief-of-staff-skill.zip
chief-of-staff.md
.claude/agents/ if it isn't there.chief-of-staff.md into it, unchanged.The two safe edits are the brief order in "When I ask for the day" (reorder the five parts to match how you read) and the horizon line (three, seven and thirty days is what works for me, and a weekly person might want one, two and eight weeks).
Teach it something once and it becomes a tool you reuse. It catches anything you've explained twice, asks two or three questions, and writes the skill file so you never explain it again.
Loading the instructions from files/claude-ai/skill-builder-instructions.md
skill-builder-skill.zip
skill-builder.md
skill-builder.md into .claude/agents/ in your project..claude/skills/formatting-meeting-notes/SKILL.md (or whatever you named it) and tells you the phrase that triggers it.The worked example inside it is meeting notes because that's the one everybody has. Swap that example for the thing you re-explain most, and the two questions it asks first stay the same.
One more pass before anything ships. It reads the whole thing, lists every claim, checks each against what's actually on the page, and hands back the gaps ranked CRITICAL, MAJOR and MINOR with a ship or hold verdict. It never fixes, it only finds.
Loading the instructions from files/claude-ai/completeness-auditor-instructions.md
completeness-auditor-skill.zip
completeness-auditor.md
completeness-auditor.md into .claude/agents/ in your project.The safe edit is the list of angles in step 2. Add the one your work always gets judged on (a board, a school district, a landlord) and it audits against that every time.
Its job is to say no. It argues the strongest case against an idea across seven tests, scores every risk, argues both sides, and closes with a verdict that either kills the idea or names exactly what has to be proven before it lives.
Loading the instructions from files/claude-ai/stress-tester-instructions.md
stress-tester-skill.zip
stress-tester.md
stress-tester.md into .claude/agents/ in your project.The safe edit is test 7, founder fit. Write two lines about what you're actually good at and what your week really looks like, and every verdict gets measured against the real you.
Credits attach to an Organization in the Console, so you paste in your Organization ID, and a Claude.ai user ID gets rejected. Each person can claim once. Anthropic runs fraud protection on this program and watches it closely, so a second attempt or a second profile gets the account flagged and at risk of being deactivated for good, and the credits team has no way to undo a bad submission. Claim it once, the right way, and the $50 is yours.
Worth saying plainly, the credits live on the API side and can't be spent on Claude.ai, so nothing you built in section 1 or 2 touched them. The friendliest way to spend them is the Workbench inside the Console, which is a chat window running on the API, and the four agents above run there too.
Workshop twoBuild your own personal assistant in Claude Code, the way Janice runs mine. This room hears the date first.
Claude Impact LabA full day building for a civic organization, in teams.
MeetupsFounders showing what they actually built, and they're free.
The date lands first in the Claude Community Chicago Discord, and the two September events on the calendar are full with waitlists open, so grab a spot on the list if either one fits.