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.

This quickstart walks you through the full bootstrap sequence from a blank ChatGPT Project to an active, loaded skill. By the end you will have the system core unpacked, the skill-adapter skill physically installed, and that skill semantically active in your session — ready to inspect and adapt external skill packages.
1

Add packages to your Project

Upload two files to your ChatGPT Project’s sources:
  • dist/GPT.SKILLS.zip — the core runtime package
  • dist/skills/skill-adapter.zip — the bundled adapter skill
Then open your Project settings and paste the full contents of CUSTOM_INSTRUCTIONS.txt into the Project instructions field. These instructions tell ChatGPT to unpack GPT.SKILLS.zip into /mnt/data/GPT.SKILLS/ at startup and immediately read the core manifest at /mnt/data/GPT.SKILLS/SYSTEM_CORE/MANIFEST.json.
At the start, unpack `GPT.SKILLS.zip` into `/mnt/data/GPT.SKILLS/`.

Read `/mnt/data/GPT.SKILLS/SYSTEM_CORE/MANIFEST.json` and follow its `load_sequence`.
Both ZIPs must be present as Project sources before you start a chat. Skill ZIPs added later require a new chat to be seen by the boot sequence.
2

Start a new chat

Open a new chat inside the Project. The Project instructions fire automatically at the start of the conversation. The core auto-unpacks GPT.SKILLS.zip into /mnt/data/GPT.SKILLS/, validates SYSTEM_CORE/MANIFEST.json, and loads the core files declared in its load_sequence:
ENTRY.md → README.md → SEMANTICS.md → OPERATIONAL_RULES.md
When the boot succeeds, the model confirms the core is active for the current session.
3

Boot the core explicitly (if needed)

If the auto-unpack did not fire — for example, because the Project instructions were not saved before the chat started — trigger core boot manually by typing:
SKILL CORE UNPACK
This command removes any previous /mnt/data/GPT.SKILLS/ tree, extracts GPT.SKILLS.zip fresh, validates the manifest, and activates the core. Running it again in the same session resets the active semantic state and clears any previously loaded skills.
4

Unpack the skill

Install skill-adapter physically into the skill directory by typing:
SKILL skill-adapter UNPACK
The system verifies that skill-adapter.zip has a flat root, a valid MANIFEST.json whose skill_name equals skill-adapter, and safe internal paths. It then extracts the archive into:
/mnt/data/GPT.SKILLS/SKILLS/skill-adapter/
The skill is now installed on disk. It is not yet active — unpack is a physical operation only.
5

Load the skill

Make skill-adapter semantically active for this session by typing:
SKILL skill-adapter LOAD
The loader reads the skill’s manifest, walks its load_sequence, and mounts each declared instruction file into the current session context. Once this command succeeds, the skill is active and the model applies its guidance to relevant tasks in this chat.
Session boundary — The active skill set exists only inside the current chat. When you open a new chat (even in the same Project), the session resets. Repeat the boot and load steps to rebuild your active skill set in the new chat. This is by design: each session starts with a clean, targeted context.

​
Next steps

Core Concepts

Understand the session-first model and what loading really means.

Session Lifecycle

Learn how sessions start, reset, and end.

Loading Skills

Explore load sequences, fallback mode, and reload behavior.

Skill Package Format

Build your own skill package from scratch.

Build docs developers (and LLMs) love

Get started for freeTalk to us