AI Context Template: Give Any Assistant a Clear Project Brief

An AI context template is a short, structured document you write once for a project and keep current as the work moves. Paste it at the start of any new session and the assistant understands the goal, the constraints, what has been decided and what needs to happen next. Without it, each new chat starts from a blank state and someone has to explain the same background again. Tools like fenbs keep that context outside any single chat app, so it is available to every connected assistant without pasting.

The template below is a starting point. Adapt the sections to fit your work and cut anything that is not useful for your project.

The template

# Project context: [project name]

## Goal
[One or two sentences. What is this project trying to achieve? What does success look like?]

## Background
[Who is involved, what the project is building or solving, and any essential history. Keep this to a short paragraph.]

## Constraints
[Budget range, timeline, technical limits, approved tools, formats or styles that are fixed. List only the ones that could affect the next task.]

## Key decisions and rules
[List the decisions that are already made and the standing rules that apply. For each one, note who decided it and why if known.]
- [Decision or rule]
- [Decision or rule]

## Current state
[What is done, what is in progress, and what has been checked. One sentence per item.]

## Next step
[The single most important thing to do next, and any information the assistant needs to begin it.]

## Open questions
[Anything that needs a decision before the next step can be completed. Who is waiting on the answer?]

What each section does

The goal section grounds everything that follows. An assistant with a clear goal can make reasonable choices when details are ambiguous. Without it, it optimizes for the task in isolation rather than for the project.

Background gives the assistant the minimum it needs to understand what it is working on. Keep this concise. A long history means the assistant must work out which parts are still relevant.

Constraints prevent work that is technically good but practically unusable. A design that ignores a fixed budget, a document in the wrong format, or code using a library that is not approved are all examples of constraints that should have been stated upfront.

Key decisions and rules are the most valuable section for ongoing projects. They prevent an assistant from reopening choices that are already made, and they apply approved rules without you restating them every session. Include the reason for each one where you have it: an assistant that understands why a rule exists applies it more sensibly.

Current state and next step are where the real session-to-session value lives. A fresh assistant reading "the form is built but delivery has not been tested" knows to treat the work as unfinished, even if the task is marked done. "Still broken" is much less useful.

Open questions flag what needs a person's decision before the work can move. When an assistant sees an open question, it should ask rather than assume. This is covered in more depth in the context engineering for AI agents guide.

The background section deserves a specific reminder: write what this assistant needs for the next task, not what someone new to the project would need to understand it fully. A full introduction to a long-running project can be a paragraph; a useful next-task brief is often shorter.

What to include and what to leave out

A useful context is short enough to read in a minute and specific enough to answer the questions a new assistant would otherwise have to ask. It is not a reference library or a project archive.

• Include: decisions that could be reversed but should not be, constraints that eliminate whole categories of approach, and the actual next step rather than a list of everything remaining.
• Exclude: background the assistant does not need for the next task, full text of documents it can access through a link, and old decisions that have been replaced.
• Keep source documents in their own tool (a shared drive, a repository, a design file). Point to them from the context with a short note on which version is current.
• Update the current state after every session. A context that still says "in progress" for something that is done is worse than no context at all.

Context rot: when the template goes stale

Context rot happens when the project moves but the context document does not. A session that reads an outdated brief may proceed confidently in the wrong direction. The AI context rot guide covers the signs and how to address them.

Build the update step into the end of each session. Before you stop, ask the assistant to note what changed in current state and update the next step. A context that costs five seconds to update at the end of a session saves ten minutes of orientation at the start of the next one.

Keeping the context where every assistant can read it

If the context lives in one chat, it is only available in that chat. If it is in a shared document, someone has to know to share it and then the assistant has to be able to read it. The gap between these two is where work gets re-explained.

How to do this with fenbs

1. Create a free account. Add your project with its goal and a short description.
2. Add the current tasks to the kanban board. Note the state of each one: what is done, what is in progress, what remains.
3. Write the standing rules and decisions in Decisions and rules. Include who decided each one.
4. Connect your AI app from Integrations. Ask it to read the project context at the start of each session.
5. Before each session ends, ask the assistant to update Where we left off and the task state. The context is current for whoever comes next.