A complete, repeatable playbook for running a hands-on Claude Code workshop for any audience size (10-50 people). Everything from pre-event attendee research to post-event cleanup.
Built from battle-tested patterns running the StellarPH x KMC Claude AI Workshop (April 10, 2026 - Cebu, Philippines - 15 attendees, 4 pods, 5 hours, 4 working apps shipped).
README.md <- you are here
FACILITATOR.md <- the full A-to-Z playbook
checklists/
pre-event.md <- 2-4 weeks before
day-of.md <- morning of the event
post-event.md <- after the session
templates/
hub/
CLAUDE.md <- hub repo AI context (templatized)
SETUP.md <- attendee setup guide (templatized)
README.md <- hub repo front page (templatized)
pod/
CLAUDE.md <- per-team AI context (templatized)
README.md <- per-team front page (templatized)
notes.md <- structured work log
guides/
claude-code-quickstart.md <- install + first commands
github-basics.md <- repos, commits, sharing
pod-kickoff.md <- pick problem, define it, build
prompt-cheatsheet.md <- copy-paste starter prompts
using-your-repo.md <- push, pull, token setup
presentation-guide.md <- NotebookLM + demo prep
tools.md <- all tools with decision guide
prompts/
facilitator/
stalled-pod.md <- when a team hasn't started
midpoint-checkin.md <- halfway pulse check format
final-push.md <- fill notes.md, prep demo
blind-spot-audit.md <- find demo-breaking bugs
rate-limit-workaround.md <- Claude web + Claude Code relay
attendee-research/
research-prompt.md <- profile attendees from reg data
profile-card.md <- per-person profile template
profiles-full.md <- full dossier with expertise map
presentations/
workshop-intro.html <- GSAP animated intro deck
collab-guide.html <- roles, handoffs, rate limits, NotebookLM
mistakes.md <- every mistake we made and how to avoid it
- Read
FACILITATOR.md- the full playbook from start to finish - Copy
templates/to your event repo, find-and-replace the{{PLACEHOLDERS}} - Use
checklists/to track progress at each phase - Use
prompts/facilitator/during the live session
| Phase | Time | What happens |
|---|---|---|
| Opening | 15 min | Setup verification, intro, pod assignments |
| Ideation | 30 min | Pick a problem, define it, assign roles |
| Build | 2.5-3 hours | Build with Claude Code, facilitator monitors |
| Midpoint | 15 min | One person per pod, 2 min each |
| Final push | 30 min | Fill notes.md, fix demo-breaking bugs |
| Presentations | 30-45 min | 3 min per pod, live demo |
- 3-4 people per pod - 5-6 causes idle members
- Each pod needs a leader who can keep things moving
- Fine-grained PATs per repo - never org-level OAuth tokens
- CLAUDE.md primes Claude Code - team context, what to build, where to get help
- notes.md IS the presentation - fill it in as you go, not at the end
- Done beats perfect - a working demo with rough edges beats a polished idea with no code
- Rate limits are normal - switch to Claude web, keep building, paste back into Claude Code
From the first run (April 10, 2026):
| Pod | Project | What they built |
|---|---|---|
| A | TindaCheck | Grocery price comparison with barcode scanner |
| B | Sakay Na! | Cebu jeepney route finder |
| C | FitAlarm | GPS fitness alarm you can't snooze |
| D | GiftMaster | Relationship intelligence PWA with AI agent |
All 4 pods shipped working apps in under 5 hours. Two were deployed live via GitHub Pages during the session.