Workshop Facilitation Protocol
Canonical multi-turn facilitation pattern for interactive PM workshops and advisors — one question per turn, progress labels, entry modes, and decision-point recommendations.
Your interactive workshop turned into a 20-question wall of text and the user bailed. Multi-turn facilitation needs structure: one question per turn, visible progress ("Q3/8"), an entry mode the user picks (Guided / Context dump / Best guess), and recommendations only at decision points. Apply this protocol on top of any workshop skill and the conversation feels guided instead of grueling.
Who it's for: Skill authors building interactive workshops, PM coaches running guided sessions, AI agent designers establishing consistent facilitation behavior
Example
"Add facilitation to my positioning workshop" → Heads-up (10 min, 8 Qs) → Entry mode picker (Guided/Context dump/Best guess) → Q1/8 with quick-select options → progress labels each turn → numbered recommendations at synthesis points only
CLAUDE.md Template
New here? 3-minute setup guide → | Already set up? Copy the template below.
npx ccplaybooks add workshop-facilitation# Workshop Facilitation Protocol
Canonical facilitation pattern for interactive skills: one question per turn, clear progress labels, adaptive recommendations at decision points, predictable interruption handling. Use as the source of truth for any guided multi-turn workshop or advisor session.
## Key Concepts
- **One-step-at-a-time** — single targeted question per turn
- **Session heads-up + entry mode** — set expectations; offer Guided / Context dump / Best guess
- **Progress visibility** — `Context Qx/8`, `Scoring Qx/5`
- **Decision-point recommendations** — enumerated options only when a choice is needed, not after every answer
- **Quick-select response options** — numbered answers + "Other (specify)" when useful
- **Flexible selection parsing** — accept `1`, `#1`, `1,3`, `1 and 3`, or custom textGet new playbooks like this one
One email a week with new Claude Code workflows. Free, like everything here.
No spam. Unsubscribe anytime.
What This Does
Defines the 12-step facilitation contract every interactive skill should follow: heads-up + entry mode picker, one-question-per-turn, progress labels, quick-select answers with multi-select parsing, decision-point recommendations only, interruption-safe flow, and best-guess assumptions tracking.
Pairs with all *-workshop and *-advisor skills as the source of truth for interaction behavior.
Quick Start
mkdir -p ~/Documents/Facilitation
mv ~/Downloads/CLAUDE.md ~/Documents/Facilitation/
cd ~/Documents/Facilitation
claude
Reference this protocol from any interactive skill's CLAUDE.md. Domain skills define content; this defines flow.
The 12-Step Contract
| Step | Behavior |
|---|---|
| 1 | Heads-up: time + question count |
| 2 | Entry mode picker (Guided / Context dump / Best guess) |
| 3-4 | One plain-language question per turn |
| 5 | Progress labels (Context Qx/N, Scoring Qx/N) |
| 6 | Follow-ups only when material |
| 7 | Numbered quick-select with "Other (specify)" + multi-select parsing |
| 8 | Recommendations ONLY at decision points |
| 9 | Accept numeric or custom; synthesize |
| 10 | Meta-question interrupt: answer + restate + resume |
| 11 | Stop/pause: halt + wait |
| 12 | End: summary + decisions + assumptions list |
Tips & Best Practices
- One question per turn. Two = a wall of text.
- Progress every turn. "Q3/8" beats silence.
- Recommendations only at decision points. After every answer = interaction drag.
- Parse loosely. Accept "1", "#1", "1,3", "1 and 3", or free text.
- Best-guess mode requires an Assumptions list. Otherwise the user can't validate.
Common Pitfalls
- Stacking multiple questions in one turn
- Numbered recommendations after every reply
- Cryptic shorthand instead of plain-language prompts
- No visible progress
- Ignoring multi-select or custom answers
- Best-guess outputs without "Assumptions to Validate"