Project context note

Canonical version: Project context note.

A project context note is a short note, one per project and one per ongoing responsibility, that an AI reads first, before it does anything else in a session. It's the concrete artifact behind Project Context Management: PCM covers everything an agent should know about a project; this note is where the agent starts reading.

Without it, the agent loads too many files, misses the right source, or asks you to explain again what you already decided. I had the same need long before AI: when I stop working on something, I must keep what I'm trying to achieve, when it should be done and what to do next (cfr Saving and restoring mental contexts). An agent forgets ALL of it between two sessions.

What goes in it

The pattern keeps showing up among adopters of the IPCRA Method. It gives each project and each cap ("casquette") a context note of five to eight lines, with five headings:

1. What it is
2. Where you stand
3. The next action
4. The decisions taken
5. The people and resources involved

His rule: short and factual. Describe, don't embellish. That note is what keeps the AI from inventing your situation.

Christian, a member of the École Cube community, keeps a _contexte.md file as the entry point of each important project or cap (where he stands, what to read, the next action, the decisions taken, the open questions). He insists it's NOT a logbook: history goes to a changelog.md, durable choices to a decisions/ folder, reusable references to Resources. Marie-Lou Girard adds one context note per project folder (what it is, for whom, the objectives).

In the Obsidian Starter Kit, the project note is the context note

In the Obsidian Starter Kit, I don't need a separate context file. Every project already has a note of the project note type, with its own metadata, lifecycle and statuses (cfr Obsidian Starter Kit - System - Action System). The same goes for activities, the kit's equivalent of IPCRA's caps: ongoing responsibilities with a cadence and a standard instead of a deadline (see Areas of responsibility and Obsidian Starter Kit - Theory - The action system). Goals, plans and tasks each get their own typed note too.

Map the five headings onto the project template and each one already has a home:

• What it is: the Objective, Success Criteria and Scope sections
• Where you stand: the status and progress properties, plus the Progress Log
• The next action: the linked task notes, each with its own status
• The decisions: the Decision Log (dated entries, never rewritten)
• People and resources: the Stakeholders and Dependencies sections, and the related_goals, related_activities and repositories properties

So the template already separates current state, history and decisions (the split Christian had to build with three files). Two more benefits come with the note type:

• It's queryable. An AI (or a Base) can list every active project, or everything waiting, without opening a single note. What each status means (done or not, success or failure, active or waiting) is declared once on the note type, so the AI reads that meaning instead of guessing it from the label (cfr Obsidian Starter Kit - System - Status System)
• It's linked up and down. A project note points to the goals it serves and the tasks it spawns, so the AI gets the "why" along with the "what"

My own agents work this way. When I tell Claude Code "let's work on X", my repository process makes it resolve X to its project note and read it (repositories, related goals, status, progress, tasks) before it touches any code.

Keep it fresh

Stale information is worse than no information (cfr Context Engineering). Update the status and the next action when you stop working, and log decisions as you take them. Then the AI, and you after a two-week break, can pick up where you left off.

References

• Nadir Imedjoubene, "IPCRA : rangez vos notes pour que l'IA les lise vraiment": https://asknadir.com/masterclass/methode-ipcra
• Christian (École Cube community), "Ma méthode pour organiser un vault Obsidian utilisable par une IA": https://community.cube.fr/forum-agents-ia/post/ma-methode-pour-organiser-un-vault-obsidian-utilisable-par-une-ia-pFZwDRMfoYyw13y
• Marie-Lou Girard, "Second cerveau" (Bravo PME, 2026-05-22): https://bravopme.substack.com/p/second-cerveau

Related

• Project Context Management (PCM)
• IPCRA Method
• Combining PKM methods into a system
• Organize information by type, not by topic
• One home per note
• Saving and restoring mental contexts
• Top of mind note
• Context Engineering
• Context File Hierarchy
• AGENTS.md (File Convention)
• Types of Context for AI Agents
• Context Reduces AI Entropy
• Knowledge-Context Pipeline
• AI-Ready Second Brain
• PKM-to-AI Readiness
• Agentic Knowledge Management (AKM)
• AI agents are only as good as the knowledge systems they operate on
• Areas of responsibility
• Obsidian Starter Kit - System - Action System
• Obsidian Starter Kit - System - Status System
• Obsidian Starter Kit - Theory - The action system
• Obsidian Starter Kit - Theory

About Sébastien

Ready to get to the next level?