Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/xxyoudeadpunkxx/canon-boundary-guard-for-gpt-project/llms.txt

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

Canon Boundary Guard is a source-bundle frame for ChatGPT Projects. It gives ChatGPT a structured protocol for keeping inspected evidence separate from chat material, agent-control instructions, model assumptions, and generated drafts — so nothing silently becomes accepted truth.

Introduction

Understand the problem Canon Boundary Guard solves and when to use it.

Quickstart

Set up the bundle in a ChatGPT Project in under five minutes.

Source Classes

Learn the five provenance layers — L0 through L3 and L1A.

Protocol Reference

Full rules for classifying sources, running the gate, and producing dossiers.

How it works

Canon Boundary Guard works by combining a Project Instructions anchor with a zipped source bundle uploaded to your ChatGPT Project. When a new session starts, ChatGPT locates the bundle, runs a mandatory Status Check, and activates the provenance-control frame before any substantive output.
1

Create and upload the bundle

Zip the canon-boundary-guard-gpt/ folder and add it to your ChatGPT Project files.
2

Paste the Project Instructions

Copy PROJECT_CUSTOM_INSTRUCTIONS.txt into the Project instructions field.
3

Start a new chat

Open a new chat inside the Project. ChatGPT will run the Status Check bootstrap automatically.
4

Work normally — the frame stays active

Continue working. The frame surfaces warnings only when a boundary is about to be crossed.

Key concepts

Operating Modes

Mode A, B, and C control how much scrutiny is applied before persistent output.

Persistence Boundary

Understand exactly what counts as a persistent write and when the gate runs.

State and Recovery

Manage SESSION_STATE and recover from lost context safely.

Scratch and Canon Zones

Learn the /mnt/data layout and the rules for promoting scratch work.
Canon Boundary Guard reduces silent promotion — it is not a hard enforcement mechanism. ChatGPT can still skip instructions or lose context. The frame is a working discipline, not a guarantee.

Build docs developers (and LLMs) love