Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/xxyoudeadpunkxx/memory-assisted-shaping/llms.txt

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

This guide walks through adding the protocol files to a ChatGPT Project and running a first shaping session. You will upload three source files, activate the protocol with a single message, and work with GPT under the operating contract defined in protocol.md. Python persistence is optional — the protocol runs without it, but notes.py gives you an append-only local log if you want it.
1

Get the files

Clone or download the repository to get the source package:
git clone https://github.com/xxyoudeadpunkxx/memory-assisted-shaping.git
The directory should contain exactly these files:
memory-assisted-shaping/
  README.md
  protocol.md
  note-io.md
  notes.py
Do not add generated session files to the source package. The files session_meta.json, session_notes.jsonl, and session_export.json are produced during use and are session-specific. Keep them out of source control — add them to .gitignore or delete them between sessions.
2

Add files to a ChatGPT Project

Open ChatGPT and create a new Project, or open an existing one. Upload the following three files to the Project’s source files:
  • protocol.md — the behavioral contract GPT follows
  • note-io.md — the I/O boundary between GPT and Python persistence
  • notes.py — the append-only Python persistence tool
These files are the source of truth GPT reads at session entry. Do not rename them — the protocol references them by name during initialization.
3

Activate the protocol

In the first message of a new conversation inside the Project, tell GPT:
Read protocol.md and use GPT-Memory-Assisted Shaping for this session.
GPT will read protocol.md and note-io.md, locate notes.py, and enter READING_ALIGNMENT mode. If Python execution is available in the environment, it will run initialization automatically:
python3 -S ./notes.py init
If Python execution is not available, GPT will state that protocol-only mode is active and that append-only persistence is not running. The shaping protocol continues to operate — memory signals will not be appended to a log, but GPT will still follow all operating states, gate handling, and synthesis discipline defined in protocol.md.
4

Shape your idea

Work normally. Describe your idea. GPT reads before shaping, shapes before synthesis, and will not produce a final artifact until you explicitly approve it.Three interaction patterns cover most sessions:Continue normal conversation to develop the idea. GPT operates in READING_ALIGNMENT until the candidate shape and real gates are clear, then moves to SHAPING to consolidate form, surface or close gates, and prepare the artifact path.Ask for a recap to recover state at any point in the session:
Recap the internal notes.
GPT will run a readback — summary first, then tail if recent context is needed — and give you a structured view of decisions, open gates, discarded paths, and artifact readiness.Approve synthesis explicitly when you are ready for the final artifact:
The shape is ready. Write the final artifact.
Implied momentum is not approval. Positive feedback is not approval. Silence is not approval. GPT will not enter SYNTHESIS until you say so. The resulting artifact will contain no process notes, discarded paths, raw memory signals, or shaping commentary.
5

Initialize Python persistence (optional)

If you have Python 3 available locally and want an append-only session log, initialize the session from the package directory:
python3 -S ./notes.py init
Expected output:
{
  "ok": true,
  "command": "init",
  "session_id": "a3f2c1d0-...",
  "path": "/path/to/memory-assisted-shaping/session_notes.jsonl"
}
The receipt confirms the session ID and the path where session_notes.jsonl will be written. From this point, every memory signal GPT emits during shaping will be appended to that log as a timestamped, checksummed record.
When state matters mid-session — especially after a long exchange or a direction change — ask GPT: "List decisions, open points, discarded paths, and artifact readiness." This triggers a structured readback without requiring a full export, and gives you a clean view of where the session actually stands before you continue shaping or approve synthesis.

​
Useful recap prompts

Copy these prompts directly into a session to recover state, check gate status, or inspect the current shape of the idea. 
Recap the internal notes.

Understand session states

Learn what GPT does in each operating state, when transitions happen, and what is forbidden in each state.

Python command reference

Full reference for every notes.py command — flags, output shapes, storage behavior, and error handling.

Build docs developers (and LLMs) love

Get started for freeTalk to us