guide
Hand your workshop prep to an AI agent
Not "AI designs your workshop" — that is a toy. This is the boring, useful version: your agent matches the activity to your brief, fetches the printable files, and builds the live measure. Three steps, real commands.
Why folders, not blog posts
An agent cannot skim a listicle and run it. Every activity here ships a SKILL.md file: frontmatter the agent matches against your brief (group size, minutes, setting, props), plus prep commands it can actually execute. The same folder's guide, debrief and assessment preset are plain markdown and JSON — no scraping, no login, nothing rendered for eyeballs only. That is the whole trick: the human reads the page, the agent reads the folder.
Step 1 — discover
Point your agent at the site's machine-readable layer: skills.json lists every skill with its description, bundle URL and sha256; catalog.json adds full folder data — group size, minutes, setting, props — as structured data. A one-line brief — "20 people, 15 minutes, no props, mixed seniority" — is enough for a matching agent to shortlist Two Truths over Paper Tower. What the agent sees is exactly what the page shows: both read the same catalog.
Step 2 — install
Each folder ships a zip bundle — one command installs it, and the unzip yields a directory named after the skill holding SKILL.md, every file it references, a manifest with per-file checksums: curl -LO https://workshopskill.com/en/activities/two-truths/files/two-truths.zip && unzip two-truths.zip. Drop that directory into your agent's skills folder — ~/.claude/skills/ (Claude Code), ~/.codex/skills/ (OpenAI Codex), ~/.qoder/skills/ (Qoder). Use the skills page for the full three-step workflow and the per-agent tutorials for exact install paths.
Step 3 — run the prep
The skill file's commands do the rest: pull the printable answer sheet and workbook, pull the debrief, and build the live pre/post measure with formlm-cli smart generate so each participant's before/after delta lands in your own workspace, auto-scored. Steps 1-3 need no account; the live measure needs a free FormLM account — and your data stays yours.
What the agent cannot do
Stand at the front of your room. The matching, the printing list, the form-building and the scoring are the agent's shift; the debrief silence — thirty seconds, nobody rescued — is yours. The folders are written so the two halves never have to meet a middleware: everything the agent prepares, you print and run exactly as the guide says.
The five steps
- Discover: point the agent at skills.json One brief line matches constraints to folders.
- Install: curl + unzip the folder's skill bundle a loader-ready directory in one command — SKILL.md, files, manifest.
- Run the prep: cards, workbook, answer sheet, live measure formlm-cli smart generate builds the auto-scored pre/post.
- You run the room The matching and scoring are the agent's shift; the debrief is yours.
Questions people ask
Which agents work with this?
Any agent that can fetch a URL and follow markdown instructions — Claude, ChatGPT with browsing, or a local CLI agent. The skills page documents the exact three-step workflow.
Is the live-measure step required?
No — every folder runs fully on paper. The online measure is for when you want automatic delta scoring or remote participants; the paper answer sheet and the JSON preset carry identical items.