Install Canon Boundary Guard for a ChatGPT Project

Installing Canon Boundary Guard means two things: uploading the zipped canon-boundary-guard-gpt/ bundle to your ChatGPT Project’s files or sources, and pasting the contents of PROJECT_CUSTOM_INSTRUCTIONS.txt into the Project Instructions field. The zip provides the frame — the skill definition, reference files, schemas, and optional scripts. The Project Instructions tell ChatGPT to locate the bundle at session start, inspect it, and use it as the operating frame before producing any substantive output. Together, the two pieces form the stable anchor that prevents chat messages, drafts, and model assumptions from being silently promoted into accepted project truth.

Project Setup (Recommended)
Manual Chat Setup

Clone or download the repository

Clone the repository or download it as a zip from GitHub:

git clone https://github.com/XxYouDeaDPunKxX/canon-boundary-guard-for-gpt-project.git

If you downloaded the full repository zip, extract it locally before continuing.

Create a zip from the bundle folder

Zip the canon-boundary-guard-gpt/ folder. This folder is the published unit — zip it as-is, preserving the internal directory structure:

canon-boundary-guard-gpt/
|-- SKILL.md
|-- references/
|   |-- gpt-project-adapter.md
|   |-- proof-of-read.md
|   |-- protocol.md
|   |-- scratch-canon.md
|   `-- state-and-recovery.md
|-- schemas/
|   |-- CANON_STATE_DELTA.schema.json
|   `-- SESSION_STATE.schema.json
`-- scripts/
    |-- artifact_fingerprint.py
    |-- extract_proof.py
    `-- validate_state.py

On macOS or Linux:

zip -r canon-boundary-guard-gpt.zip canon-boundary-guard-gpt/

On Windows, right-click the folder and choose Compress to ZIP file.

Create a ChatGPT Project

In ChatGPT, open the sidebar and create a new Project. Give it a name relevant to your work. Projects provide the stable container that gives the frame persistent continuity across sessions.

Add the zip to Project Sources or files

Inside the Project, open Project Settings (or the files/sources panel) and upload canon-boundary-guard-gpt.zip. ChatGPT will store this as a Project Source that persists across all chats in the Project.

Paste the Project Instructions

Still in Project Settings, open the Project Instructions field. Copy the full, unmodified contents of PROJECT_CUSTOM_INSTRUCTIONS.txt from the repository root and paste them verbatim. Do not paraphrase or shorten the text — the exact wording is the instruction anchor.

Start a new chat inside the Project

Open a new chat within the Project. ChatGPT should immediately run Status Check: locate the bundle, extract if needed, inspect the bootstrap sources, and confirm the frame is active before producing any substantive output.

Upload the zip in a normal ChatGPT conversation

In any standard ChatGPT chat (outside a Project), use the file attachment button to upload the canon-boundary-guard-gpt.zip bundle directly into the conversation.

Ask ChatGPT to inspect the bundle and activate the frame

Send a message instructing ChatGPT to treat the uploaded bundle as the active provenance-control frame for the session. For example:

Please inspect the uploaded canon-boundary-guard-gpt bundle and use it
as the active Canon Boundary Guard frame for this session.

ChatGPT should locate the zip in /mnt/data, perform source-staged extraction, inspect SKILL.md and the reference files, and confirm the frame is active.

Manual chat setup is less stable than a Project. When the frame is anchored only inside a conversation, it does not persist across new chats. Every new conversation requires re-uploading the bundle and re-activating the frame. For sessions where continuity and reliable bootstrap matter, use the Project setup instead.

Bundle structure

The canon-boundary-guard-gpt/ folder is the complete published unit. The tree below shows every file included when you zip it for upload:

canon-boundary-guard-gpt/
|-- SKILL.md
|-- references/
|   |-- gpt-project-adapter.md
|   |-- proof-of-read.md
|   |-- protocol.md
|   |-- scratch-canon.md
|   `-- state-and-recovery.md
|-- schemas/
|   |-- CANON_STATE_DELTA.schema.json
|   `-- SESSION_STATE.schema.json
`-- scripts/
    |-- artifact_fingerprint.py
    |-- extract_proof.py
    `-- validate_state.py

SKILL.md is the primary skill definition and the first file ChatGPT inspects during bootstrap. The references/ files define provenance layers, the GPT Project persistence boundary, state and recovery rules, scratch-to-canon promotion rules, and proof-of-read requirements. The schemas/ files are used to validate session state and snapshot deltas. The scripts/ files are optional Python helpers for mechanical validation — they are not hooks and do not make provenance decisions.

A ChatGPT Project is the recommended container for running Canon Boundary Guard. Uploading the bundle as a Project Source means it is available at the start of every chat inside the Project, giving the frame a stable and persistent home. Manual chat use is possible but provides weaker continuity.

Overview

Setup

Using the Frame

Validators & Scripts

Bundle structure

Build docs developers (and LLMs) love

Overview

Setup

Using the Frame

Validators & Scripts

Documentation Index

​Bundle structure

Build docs developers (and LLMs) love

Bundle structure