If your team uses the same prompts over and over in Copilot in SharePoint and gets slightly different answers every time, the missing piece is usually one Markdown file. SHAREPOINT.md sits in your site's Agent Assets library and loads automatically into every chat. One file, one site, every conversation starts with the same context.
This is the walkthrough of what to put in the file and a live demo of it driving a sequential rename of eight policy files. Six minutes of video, one Markdown file, no prompt engineering.
What SHAREPOINT.md actually is
A single Markdown file you place at the root of your site's Agent Assets library. The path is /Agent Assets/SHAREPOINT.md, uppercase filename, plain Markdown body.
Copilot in SharePoint discovers it automatically when you open chat on the site. It loads the contents into the conversation as background context. Everything the user types after that is interpreted with the SHAREPOINT.md content already in mind.
The user does not have to attach the file. The user does not have to mention the file. The user does not even need to know the file exists. The site owner who wrote the file is doing the work for every team member who runs a chat on the site.
This is the What to Know capability Microsoft named in the April 2026 Tech Community blog. The other two capabilities are Skills and content generation, covered in the three capabilities article.
What goes inside the file
Eight sections cover most useful SHAREPOINT.md files. The template at copilotinsharepoint.com/sharepoint-md-template.md has all of them ready to fill in.
Site identity. Name, URL, purpose in one sentence, primary audience, site owner.
What this site is for. A short paragraph on the work this site supports. AI uses this to decide whether a request is in scope.
Library map. Every library on the site, with path, what it contains, naming convention, and any rules. This is the biggest single value-add. Without it, every chat has to discover where things live.
List map. Same idea for lists. Path, purpose, key columns.
Naming conventions. Document-level and list-level naming rules, with examples.
Voice and document standards. House style (Australian English, US English), reading level, tone, banned words, required phrases.
Rules for AI. Explicit rules to follow when generating or editing content on this site. Save new policies to /Policies. Never publish outside the site without confirmation. Flag missing metadata rather than overwriting.
Glossary. Org-specific or site-specific terminology that AI should use consistently.
Written for AI, not for humans. Prefer tables, lists, and short labelled sections. Avoid long prose paragraphs that AI has to parse.
Live demo: renaming eight policies sequentially
In the video, the SHAREPOINT.md file on a people and culture site includes a naming convention for the policies library: every policy should be prefixed with POL plus a three-digit sequential number.
The user opens Copilot in SharePoint on the site and asks, "according to the naming conventions, can you please rename all of these policies in this library?"
Three things happen automatically.
First, AI in SharePoint loads the SHAREPOINT.md file. It now knows the naming convention without anyone telling it explicitly.
Second, it locates the policies library based on the library map in SHAREPOINT.md. It does not need to be told which library to act on.
Third, it applies the naming convention sequentially. POL001. POL002. POL003. Through to POL008. All eight policies renamed in one operation, exactly to the standard the SHAREPOINT.md file specified.
The user prompt was twelve words long. The SHAREPOINT.md file did the rest.
Why this changes how the team works
Without a SHAREPOINT.md file, every chat on the site starts from zero. The user has to explain the naming conventions, the library structure, the rules to follow. The senior team member who knows all this either explains it in every prompt or accepts inconsistent output.
With a SHAREPOINT.md file, the standards are in the site. The senior team member's institutional knowledge is encoded once. New team members inherit it on day one. A team member who joins the company three months from now opens chat, types a request, and the output respects the site's standards because the file taught AI in SharePoint about them automatically.
The pattern is the same as a good onboarding document for a new starter. You do not re-explain the team's conventions on every meeting. You write the conventions down once and the new starter inherits them. SHAREPOINT.md is that document, written for AI.
Common mistakes when writing SHAREPOINT.md
Three patterns I see in client tenants when the file underperforms.
Too long. A SHAREPOINT.md over 4000 words starts to load slowly or partially. Keep the file focused on the rules and conventions that affect AI behaviour. The full site documentation belongs in a separate site or a knowledge base, not in this file.
Written for humans, not AI. Long flowing prose paragraphs are harder for AI to parse reliably than structured tables and short labelled sections. Write the file like a configuration document, not an internal blog post.
Out of date. When the library structure changes, the naming convention changes, or the team's standards evolve, SHAREPOINT.md needs to keep up. A stale file produces stale outputs. Add a maintenance entry to your team's change-control process.
Stuffing Skills logic into SHAREPOINT.md. This is the third capability mistake. SHAREPOINT.md is for passive context that applies to every chat. Multi-step workflows belong in Skills, not in the context file. If you find yourself describing a process step-by-step in SHAREPOINT.md, it should be a Skill instead.
What to do this week
Three actions.
-
Activate the Agent Assets feature on the site where you have the most active Copilot in SharePoint usage. Site settings, Site collection features, find Agent Assets, activate.
-
Download the template at copilotinsharepoint.com/sharepoint-md-template.md. It is annotated and has every section ready to fill in.
-
Spend an hour with the template. Fill in the site identity, the library map, and the naming conventions for one or two libraries you use most. Upload as SHAREPOINT.md to the root of Agent Assets. Test on the next chat.
The payback is on the very first chat after the file is in place. The team's most repeated context, written down once, applied automatically every time.
For organisations rolling Copilot in SharePoint out across multiple teams, the Copilot in SharePoint Immersive is a one-day workshop where we write the SHAREPOINT.md file for one of your most active sites together. By morning tea your site has the file in place. The afternoon goes into Skills. By the end of the day your team is using all three capabilities, not just talking about them.
