onboarding-trial

Dispatches randomized trial assignments and manages the onboarding file queue for experiment sessions.

Updated Apr 11, 2026
One-click install
npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill onboarding-trial-gozonerd
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: onboarding-trial
Source: https://github.com/gozonerd/shadow-ai-assessment/tree/main/.claude/skills/onboarding-trial
Command: npx skills add https://github.com/gozonerd/shadow-ai-assessment --skill onboarding-trial-gozonerd

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Running a randomized controlled trial across many AI sessions requires consistent session initialization, blinded condition assignment, and reliable queue management. This Skill automates the session-opening protocol so each experiment session starts with the correct trial assignment without manual file handling. ## Core Features & Use Cases - Trial Dispatch: Reads the single assignment file in the next-trial-assignment folder and follows its onboarding instructions silently. - Queue Management: Moves completed assignments to used-trial-assignments and promotes the lowest-numbered file from unused-trial-assignments, keeping the queue sequential. - Blinding Enforcement: Never reveals the assigned condition or onboarding content to the user, reporting only the 4-digit trial number. - Use Case: A researcher running a Growth Mindset Onboarding RCT opens each Claude Code session with this Skill, which assigns the next trial, completes the onboarding protocol, and confirms readiness with the trial number. ## Quick Start Start the session by invoking the onboarding trial skill to receive my trial assignment and confirm the trial number.

Frequently Asked Questions about onboarding-trial

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

FAQPage Schema
How do I run a randomized controlled trial across multiple AI sessions?

Use a session-opening protocol that reads one trial assignment file per session, follows its onboarding instructions, and rotates the file queue. This Skill automates that flow by dispatching the next assignment and advancing the queue after each session.

How to keep experimental conditions blinded from the user during a trial?

Blinding is enforced by processing onboarding instructions silently and reporting only the trial number. The Skill never reveals the assigned condition, the onboarding content, or the sealed condition-key folder to the user.

What happens if the trial assignment folder is empty or has multiple files?

The protocol stops immediately and reports the specific problem, such as an empty next-trial-assignment folder or multiple files present. It does not guess, improvise, or proceed until the queue is manually corrected.

How does the trial file queue stay in sequence between sessions?

After each session, the completed assignment moves to used-trial-assignments and the lowest-numbered file in unused-trial-assignments moves into next-trial-assignment. Verification confirms exactly one file remains in the next folder.

When should this onboarding dispatch approach not be used?

It is unsuitable outside the specific RCT folder structure it expects, since it depends on next, used, and unused assignment directories. General session onboarding without trial blinding or queue management does not need this protocol.