# Fac Role Play Practice Kit

> Builds the practice for a conversation people must learn, producing four to six learning points, a good and a poor model of the same conversation, role cards for trios, rounds that move to cases participants bring, and observer notes that go only to the learner. Use for "run fac-role-play-practice-kit", "role play scenarios for training", "role play without the cringe", "difficult conversation practice", "feedback conversation practice", "behaviour modelling", "observer notes", "people hate role play", part of the Claude for Trainers and Facilitators Pack by 90n.

- Skill: `polar-bear-org/fac-role-play-practice-kit` (Agent Skill)
- Install (CLI): `npx get-skill add polar-bear-org/fac-role-play-practice-kit`
- Raw SKILL.md: https://api.skillmd.com/api/skills/polar-bear-org/fac-role-play-practice-kit/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: polar-bear-org (https://skillmd.com/u/polar-bear-org)
- Updated: 2026-11-14
- Page: https://skillmd.com/skills/polar-bear-org/fac-role-play-practice-kit

---


# Role-Play Practice Kit
## When To Use
Role play is the part everyone dreads, and the part the skill most depends on. Use this when the thing people must be able to do is a live conversation (giving feedback, saying no to a client, a first coaching talk) and they need to try it with each other before they try it for real. It answers: what exactly do people practise, what does good look like, and how does anyone take part without being put on show?

## When Not To Use
For group methods that are not a practised conversation, use Group Activity Bank. For the full question sequence after any activity, use Activity Debrief Questions; this kit keeps only the few questions specific to the practice.

## Inputs
- The conversation to practise, the learners' role, and what they should do at work afterwards.
- Real situations from the job, reworded: what the other person tends to say, and where it usually goes wrong. Also the headcount and the minutes you give to practice.
If you have none of this, I start from the conversation and the role and mark the output as a first draft.

## Approach
Behaviour modelling training, as reviewed in the 2005 meta-analysis of behaviour modelling training by Taylor, Russ-Eft and Chan: clear learning points, mixed good and poor models, scenarios trainees bring and goal setting are the features it links to skills carrying over, and it reports that effects shrink from the room to the job. The judgment is to lower the cost of taking part, not the difficulty of the conversation. The failure this prevents: one pair performing at the front while everyone else watches and hopes not to be next.

## Workflow
1. Ask at most three questions: which conversation and with whom, what goes wrong in it today, and how many people with how many minutes for practice?
2. Write four to six learning points as short rules a person can hold mid-conversation ("name the problem before the fix"). If a point needs a paragraph, it is content, not a learning point.
3. Build mixed models: one good and one poor example of the same conversation, each marked with the learning points it shows or misses, from your real situations and labelled as written examples. Then you read both aloud and ask "would someone in this job recognise this conversation?" before using them.
4. Set up trios: learner, other party, observer. Each role card gives the situation, what the other party wants and how hard to push. A card that makes the other party a villain teaches nothing.
5. Plan the rounds: round one on a prepared case, later rounds on cases participants bring, reworded and with no names. Roles rotate.
6. Offer three ways to take part, chosen by the participant: play the learner, play the other party, or observe and give notes. Nobody performs in front of the whole room unless they offer.
7. Write the observer note against the learning points only (saw, worked, one thing to try), handed to the learner and never collected. Add two or three practice debrief questions, and close with each learner setting their own goal for the job.

## Output Format
```markdown
# Role-Play Practice Kit
Conversation: [what is practised] | Learners: [role] | At work this means: [what people will do]
## Learning points
1. [short rule] 2. [short rule] 3. [short rule] 4. [short rule]
## Two models of the same conversation (written examples, not real quotes)
| Model | Outline | Learning points shown or missed |
|---|---|---|
| Good | [outline] | [points shown] |
| Poor | [outline] | [points missed] |

Read aloud by [trainer] on [date]. Would someone in this job recognise this conversation? [yes / changed to]
## Role cards for trios
| Round | Case | Learner's situation | Other party wants, and pushes this hard | Observer watches for |
|---|---|---|---|---|
| 1 | Prepared | [situation] | [wants, how hard] | [learning points] |
| [later round] | Brought by participants, reworded, no names | [situation] | [wants, how hard] | [learning points] |

Rounds: [number] of [minutes] each, roles rotate | Ways to take part, each person chooses: learner / other party / observer
## Observer note (handed to the learner, never collected)
Saw: [what happened] | Worked: [learning point] | One thing to try: [one thing]
## Practice debrief and own goal
Questions: [question] / [question] / [question] | Rounds and minutes: set by the trainer
My goal at work (kept by the learner): [what I will try, in which situation]
## Decision
[Trainer] confirms the learning points and both models by [date], after reading them aloud to [someone who does this job].
```

## Done When
- Each learning point is short enough to recall mid-conversation, both models show the same conversation, and you have read them aloud and judged that someone in the job would recognise it.
- At least one round runs on cases participants bring, and every person has three ways to take part.
- Every learner leaves with a goal they set themselves.

## Quality Bar
- Role cards sound like real people, built from your inputs; no invented quote is presented as real.
- Observer notes describe what happened against the learning points, never the person.
- No scoring sheet, no rating scale, no summary of who did well; a request for one is refused.
- Observer notes go only to the learner; no role play is scored, ranked or reported, and nobody performs in front of the room unwilling.

## Next
Run fac-virtual-session-redesign (Virtual Session Redesign) to rebuild any of this that has to run on a video call.

## About the makers

This pack is made by 90n ([90n.ai](https://90n.ai)): AI-native ways of working and tools for Product, Marketing and HR teams, built by ex-McKinsey founders. We map where time and money leak, redesign the work around AI and stay until your people run it themselves, and we run our own company on Claude. If your team has outgrown the self-serve version, message Pauline (linkedin.com/in/paulinebertry).
