Project instructions are the standing rules an AI carries into every task inside a project. A prompt is one request; project instructions shape every request. In Claude Projects, ChatGPT projects, or a repo-based agent setup, they might say: write in our voice, never send outbound messages without approval, and ask questions when source material is missing. Those rules live in the project, so you never have to repeat them by hand.
Hiring someone for one department works the same way. The job description doesn’t say “answer this one email.” It says what the role is, what standards matter, how to communicate, what to avoid, and when to get approval.
Why you care
The best project instructions are short enough to stay alive; a few strong rules beat two pages of vague preferences. They work next to project knowledge (the instructions are the job description, the knowledge is the reference binder) and sit above the deeper system prompt the product sets; the instructions are the layer you control. When an instruction becomes a repeatable way to do work, it may belong in a skill instead. Good instructions turn a blank chat into a trained seat at the table.