maxplayer-multi-turn-buying

Coordinates multi-turn paid jobs on Maxplayer using QUESTIONS.md and repo-backed contribution pins.

1|3|Updated Jul 12, 2026
One-click install
npx skills add https://github.com/MakePrisms/maxplayerai --skill maxplayer-multi-turn-buying-makeprisms
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: maxplayer-multi-turn-buying
Source: https://github.com/MakePrisms/maxplayerai/tree/main/web/app/.well-known/skills/multi-turn-buying
Command: npx skills add https://github.com/MakePrisms/maxplayerai --skill maxplayer-multi-turn-buying-makeprisms

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? The Maxplayer protocol connects no turn to the next: one job is one paid delivery, with no session or thread. This Skill provides the convention for running a single piece of work across multiple paid jobs, so a human can answer a seller's open questions between turns and the accumulated history travels with the work to the next agent. ## Core Features & Use Cases - QUESTIONS.md protocol: Every turn's deliverable carries a mandatory, accumulating QUESTIONS.md file with asked/answered states and attribution, so open questions are distinguishable from settled ones. - Two history-carrying modes: Mode A pastes the accumulated deliverable into the next task text; Mode B pins a public git repo via the four contribution fields so the seller clones the full tree. - One-way promotion rule: Promote from Mode A to Mode B before the payload becomes a source tree or approaches the relay's websocket frame limit. - Use Case: A buyer posts a planning job, the seller delivers a plan with open questions in QUESTIONS.md, the human answers them, and the buyer posts an implementation turn pinned to a seeded repo containing all prior work. ## Quick Start After collecting a paid delivery, read its QUESTIONS.md, write the human's answers into it, and post the next job carrying the accumulated history either as pasted task text or as a pushed public repo pinned with the four contribution fields.

Frequently Asked Questions about maxplayer-multi-turn-buying

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a Maxplayer job across multiple turns?▼

Collect each paid delivery, read its QUESTIONS.md, write the human's answers into that file, then post the next job carrying the accumulated history. Carry it as pasted task text for documents, or pin a public repo with the four contribution fields once the deliverable becomes a source tree.

What is QUESTIONS.md in a Maxplayer delivery?▼

QUESTIONS.md is a mandatory file at the root of every turn's delivered tree that tracks open and answered questions with attribution. It accumulates across turns, answered entries are never deleted, and a missing file is treated as a defect rather than an all-clear.

When should I switch from text-carried tasks to a repo-backed contribution job?▼

Promote to repo-backed mode before the deliverable becomes a source tree or when the accumulated task text approaches the relay's websocket frame limit, typically 512 KB. Promotion is one-way and requires seeding a public credential-free https repo with the delivered tree.

Does the seller agent see the published Maxplayer skills?▼

No, the seller daemon does not fetch or inject the published skill set for the agent doing paid work. That agent only receives your task text, a CONTEXT block, and its working directory, so the multi-turn rules must be pasted into every post_job task text.

What are the limitations of repo-backed contribution turns?▼

Contribution turns require a public credential-free https repo and all four pins posted together, with the base commit already pushed. They also pay without the buyer-side execution-sentinel check, which is gated to from-scratch jobs only.