Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/XxYouDeaDPunKxX/ChatGPT-SKILL-SYSTEM/llms.txt

Use this file to discover all available pages before exploring further.

The GPT Project Skill System brings a structured skill-package workflow to ChatGPT Projects. Instead of copying instructions by hand into every new chat, you package reusable skills as ZIP files, upload them to a Project, and load them explicitly when you need them. Skills are active only for the current session — no permanent state, no automatic routing, no hooks.

Quickstart

Boot the core and load your first skill in a new ChatGPT Project in minutes.

Core Concepts

Understand sessions, skill packages, lifecycle commands, and the mount model.

Build a Skill

Learn the ZIP package format, manifest fields, and file categories.

Command Reference

Full reference for SKILL CORE UNPACK, SKILL UNPACK, and SKILL LOAD.

How it works

The system is built around three explicit steps you run inside a ChatGPT Project chat:
1

Upload packages to your Project

Add dist/GPT.SKILLS.zip and any skill ZIPs (e.g. dist/skills/skill-adapter.zip) as Project sources. Add CUSTOM_INSTRUCTIONS.txt to your Project instructions.
2

Boot the core

Start a new chat and run SKILL CORE UNPACK. This extracts the runtime core from GPT.SKILLS.zip into /mnt/data/GPT.SKILLS/ and activates it for the session.
3

Load the skills you need

Run SKILL skill-adapter UNPACK then SKILL skill-adapter LOAD (or any other skill). The skill’s declared load_sequence files are read into the session context and the skill becomes active.
The active skill set exists only inside the current chat session. When your work focus changes, open a new chat and repeat the boot/load cycle with the skills relevant to your next task.

What this is not

The system intentionally excludes features that would add complexity without benefit:
  • No permanent skill installation or auto-discovery
  • No hooks, watchers, or daemons
  • No dependency installation (pip install is not supported)
  • No marketplace or automatic routing
  • No persistent runtime state between sessions

Build docs developers (and LLMs) love