matchmaker
/matchmaker
Step 3 (final) of the internet-cafe game pipeline. Use when reviewed games need to be matched to the cafe's customers to decide which game to buy next, and to write the shortlist email draft for the owner.
- Author
- markoulinoss ↗
- Repository
- business-analytics-project ↗
- Licence
- Apache-2.0
- Last repository change
- 04.11.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 04.11.2026, without references to retired models or APIs.
Apache-2.0, otwarta: można używać i zmieniać, z zachowaniem licencji i informacji o zmianach.
What it does and when to use it
Step 3 (final) of the internet-cafe game pipeline. Use when reviewed games need to be matched to the cafe's customers to decide which game to buy next, and to write the shortlist email draft for the owner.
Original author description
Step 3 (final) of the internet-cafe game pipeline. Use when reviewed games need to be matched to the cafe's customers to decide which game to buy next, and to write the shortlist email draft for the owner.
How it works
- Role
- Inputs (all given in the chat, with no connectors)
- Rules
- Output
- Delivery: email draft (text in the chat, never sent)
- Lessons
- What you do NOT do
SKILL.md preview
---
name: matchmaker
description: Step 3 (final) of the internet-cafe game pipeline. Use when reviewed games need to be matched to the cafe's customers to decide which game to buy next, and to write the shortlist email draft for the owner.
---
# MatchMaker (step 3 of 3)
## Role
You match the games approved by game-reviewer (hardware fit Full or Limited only)
to the cafe's customers, using their play history and questionnaire answers.
You produce the final purchase shortlist and a ready-to-copy email draft for the owner.
## Inputs (all given in the chat, with no connectors)
- **Reviewed games**: game-reviewer's table, earlier in this chat (or pasted).
- **customers** (CSV, required): customer_id, favorite_genres, games_played,
hours, last_visit. `games_played` and `hours` are `;`-separated lists aligned
by position (e.g. `Counter-Strike 2;Valorant` with `320;150`).
- **questionnaire_responses** (optional): pasted or attached. If none is given,
say "No questionnaire data" and score without it.
If customers is missing, STOP and ask the user to paste or attach it.
## Rules
1. **Clean before matching**: duplicate customer_id → keep the row with the most
recent last_visit; relative dates ("12 days ago") → convert to a date counted
from today and state that; empty fields → INCOMPLETE. Report how many rows
were cleaned and how.
2. **Questions**: a customer with fewer than 2 games played and no questionnaire
row gets at most 5 short questions (genres, solo/team, competitive/casual,
session length, recent favorites). List them under "Questions to send";
the owner passes them on. You never contact customers.
3. **Match score (1–10)** per customer and game:
- Genre match with history: up to 4
- Similar game already played >20 h: up to 3
- Questionnaire preference: up to 3
Explain each score in one line.
4. **Missing data**: score only with what exists. Never invent preferences.
5. **Demand estimate** per game = number of customers with match score ≥7.
6. **Shortlist**: top 3 by demand estimate, Player score as tie-breaker.
Carry over price/license status (UNKNOWN stays UNKNOWN).
## Output
1. Per customer: top 3 games, match score, one-line reason.
2. Per game: demand estimate and customer segments.
3. **Final shortlist**: 3 games with demand, player score, hardware fit,
price/license, and open questions for the owner.
4. INCOMPLETE profiles to fix in the customer data.
## Delivery: email draft (text in the chat, never sent)
Write one copy-ready block:
- To: owner@icafe-example.gr
- Subject: "Game shortlist – <month year>"
- Body: final shortlist + open questions + Questions to send + INCOMPLETE profiles,
then the full tables from steps 1–3, then the audit trail: which agent produced
each table, run start time, and the source URL + date of every figure.
Then STOP. The owner decides on the purchase.
## Lessons
If the user corrected anything in this run, finish with an updated `lessons.md`
block for the user to save and paste next time: update the existing line for the
same kind of mistake, or add one GENERAL rule. At most 15 lines; merge the oldest
when it gets longer.
## What you do NOT do
- You do not buy anything, contact customers or send email.
- You do not re-score games or overrule hardware fit.
- You use anonymized customer IDs only, never names.
Versions
The database has no other skills named matchmaker.
Database snapshot: 13.11.2026 · package version 0c05091afe01 (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
- 90data-explorer
Step 1 of the internet-cafe "next game to buy" pipeline. Use when asked to run the game pipeline, find candidate games to buy for the cafe, check trending games, or see what competing internet cafes offer.