Guide Library

Agent Workshop

Agent Workshop - Connect Your Agent to Your Information Inside ChatGPT

Week 3Agent Workshop

  • 20–30 minGuided walkthrough

Use a desktop browser for this workshop. Put your Safety & Security Rules in My Agent, your ChatGPT Project, with a version number, and tell your agent to follow them. Then learn what a plugin can reach, who decides which ones you can use and what ChatGPT lets you limit. You finish with a way to give your agent your meetings.

Watch the video walkthrough in StormWind Campus (sign in with your bootcamp login).

What You Need for This Guide

  • A desktop browser.
  • Your approved ChatGPT work account with Project and file-upload access.
  • Your My Agent Project from the last two workshops, with your and style guide in it.
  • Your saved My Safety & Security Rules from this week's session.

Give Your Agent Your Voice Inside ChatGPT

Build Your Safety & Security Rules

Choose the tool you’re allowed to use

Your approved tool

The tool your company approved.

Pick your approved tool above to see the steps and screenshots for it.

Start guide
7 steps · At your own pace
About saved progress

No account needed. Checkmarks save on this browser and device when browser storage is available. If storage is blocked or full, new checkmarks last until you reload or close this page. Progress doesn’t sync to other browsers or devices. Clearing this site’s browser data removes it, and private browsing may not keep it after you close the window.

Standard Data rules for AI at work

Data Rules

Data rules for AI at work

Check with your IT Team before you paste any work data into an AI tool. Avoid names, account numbers, customer data, or anything sensitive without explicit permission from IT. If you aren't sure, redact anything sensitive and replace it with generic information. AI makes mistakes, make sure to fact check your agents outputs even if they sound right.

Before you start

  • Check with your IT teamAsk what your company allows before you put anything from work into an AI tool. Your company policy is the one that counts, and it is the one this library cannot see.
  • Use the tool your company approvedChatGPT, Claude, Microsoft 365 Copilot Chat and Gemini are all covered here. Use the one your company approved, not the one you like best.
  • Sign in with your work accountA work account keeps your chats inside the agreement your company signed. A personal account in the same product sits outside it.
  • Connected appsAn email or file connection your IT team approved can read the names and addresses in what it opens. Everything else under “Never put this into an AI tool” still applies.
  • No approved account yetFollow the steps with made-up example text, or on paper, until IT gives you one.

Never put this into an AI tool

  • Customer or personal detailsNo names, email addresses, phone numbers, home addresses or account numbers. That covers colleagues as well as customers.
  • Regulated recordsHealth, financial, HR and legal records stay in the systems built to hold them, unless IT has told you in writing that this tool may see them.
  • Passwords, keys and access codesNo credentials of any kind, yours or anyone else’s, and nothing that grants access to a system.
  • Anything under an NDA or not yet publicUnreleased plans, contracts, pricing and source code need clearance before they go near a chat window.
  • Anything your company policy prohibitsYour company’s rules win over anything in this library. If policy keeps something out of AI tools, it stays out, whatever a guide suggests.

Do this instead

  • Swap real details for placeholdersWrite “Customer A”, “[client name]”, “$X”. You get the same quality of answer, and you put the real details back yourself.
  • Describe the situation rather than pasting the documentMost of what an agent needs is the shape of the problem, not the file it arrived in.
  • Trim a file before you upload or paste itTake out the identifying columns and pages first, or work from a short extract.

Check the answer before you use it

  • Read it as if a colleague had drafted itAn agent states wrong things with complete confidence. Your name goes on the work, so the check is yours.
  • Say where the agent helpedFollow your company rule on disclosing AI assistance, especially in client-facing or formal work.

When you are not sure

  • Ask before you paste or upload, not afterA question to your IT team takes a minute. Cleaning up after a data incident takes a great deal longer.
  • When in doubt, leave it outIf you would not put it in an email to someone outside the company, keep it out of the chat.
Open the full page
Download Guide

Step

Add your rules to My Agent and number them

Your Safety & Security Rules from this week's session tell your agent how to handle information, check its work and flag what needs a person. In My Agent's Sources, every chat in the Project follows them, and a version number keeps a record of every change.

This builds on the last two workshops, Give Your Agent a Home Inside ChatGPT and Give Your Agent Your Voice Inside ChatGPT, and on the rules you saved with Build Your Safety & Security Rules. Do those first if you haven't.

If your account doesn't offer Projects or file uploads, watch the video and keep attaching or pasting your , style guide and rules into each chat, or ask IT about access. Skip the parts that save files or instructions in My Agent, and send each card in a new chat with your Agent Profile, style guide and newest rules file attached or pasted in.

  1. In My Agent Files, make a folder called Safety Rules beside your Agent Profile and style guide folders, and move your newest My Safety & Security Rules file into it.

  2. In ChatGPT, open your My Agent Project and select Sources, then Add sources, then Upload, and choose your rules file.

  3. Start a new chat in the New chat in My Agent box and send this card. It adds a RULES VERSIONS section, changes nothing else and gives you back the whole file as the next version.

    Set up versioning for your rules (no change)
    I want to set up versioning for my Safety & Security Rules, so I can track every change.

Change to make: None, set up versioning only

Put any change in my rules file, not your built-in memory, so my file stays my master copy. Unless a rule conflicts, do all of this in one reply:

1. If I gave a change, put it in the section it fits best, such as My Organization-Specific Rules. Use my words, and don't add rules I didn't give you. If it would weaken or contradict a rule already there, stop before writing anything: quote both and ask me which one to keep.
2. If my rules have no RULES VERSIONS section, add the one below to the end of the file, word for word. If the file has no version number in its title, version line or filename, treat the current file as Version 1. Otherwise continue from its current number. Do not restart numbering.
3. Follow RULES VERSIONS to return the entire updated file, with the new version number in both its title and its filename. Keep every other section and rule word for word, including Rule Priority and anything I've added to a rule.
4. In the What changed table, list my change, if I gave one, and say whether you added RULES VERSIONS. Do not claim a section was added if it was already there.
5. End by walking me through saving it, as RULES VERSIONS describes.

RULES VERSIONS

The rules title, the version line under it, and the filename all show the same version number. Use simple whole numbers: v1, v2, v3, and so on. The title uses readable spaces, like My Safety & Security Rules v2. The filename uses hyphens and no spaces, like My-Safety-and-Security-Rules-v2.txt.

- If I ask which version you're using, give that line's number and date.
- If I ask you to update these rules, create the next version:
  1. Raise the current version number by one. Use that new number in the title and in the version line below it, like Version 2 | Updated October 20, 2026. These are examples: use the actual next number and today's date. If you can't see today's date, write DATE NEEDED.
  2. Apart from the title, version line and version history, change only what I asked. Keep every other rule and section word for word, in the same order.
  3. Add a one-line summary to the top of VERSION HISTORY. Keep the five newest lines.
  4. Give me the entire updated rules file as one downloadable plain-text file named My-Safety-and-Security-Rules-vNUMBER.txt, replacing NUMBER with the new version number. Use only plain letters, digits, hyphens and the dot before txt. Include every section and all unchanged content. If you can't create files, put the entire file in one code block and give me the exact filename above it.
  5. Show what changed in a table: Section, What changed, Why.
  6. Walk me through saving it: save the new file in the same folder as my earlier versions, keeping every older file. Then replace the rules in your saved setup the same way I first added them, start a new chat, and ask which version you're using.
- Never weaken, shorten or remove a rule unless I ask for that exact change. If a change would weaken or contradict a rule already here, quote both and ask me which one to keep before you write the new version.
- If I tell you something new that belongs in these rules, offer a new version. Wait for my yes.
- Never overwrite an earlier version or reuse its number.
- You can't change your own saved setup, so don't say my rules are updated until I've replaced the file.

VERSION HISTORY
- Version 1: Original rules.

  4. Check the reply before you save anything. The What changed table should say RULES VERSIONS was added, the new file's title, version line and filename should all show the next version number, v2 for most people, and every rule you wrote should still be there. If something's wrong, ask in the same chat for a complete corrected file.

Step

Save the new version and replace your rules in My Agent

Saving the new version on your computer doesn't change My Agent. Until you replace the file there, every chat in the Project still reads the earlier one.

  1. Download the new file from ChatGPT's reply. Save it in your Safety Rules folder under the filename in the reply, such as My-Safety-and-Security-Rules-v2.txt, and keep the earlier file beside it as your record.

    If ChatGPT gives you text instead of a file, copy all of it into a new plain-text file with that exact filename.

  2. In My Agent, select Sources, point at your earlier rules file and open its ⋯ menu to remove it from the Project. Then select Add sources, then Upload, and choose the new file. Sources should now list three files: your , your style guide and your new rules.

Step

Tell My Agent to follow your rules

This card replaces last week's instructions. It keeps every line and adds one, for your Safety & Security Rules: follow them in every chat. The details stay in your files.

  1. Open the three-dot menu beside Share and choose Project settings. The Instructions box is small, so click into it and select all of last week's text, then paste the card in its place and select Save. If you added lines of your own last week, add them to the new card before you save. Reopen Project settings to check the whole card was saved, through the last line about master files.

    Project instructions with your rules
    You are my work agent. You help me with my day-to-day work.

You have these reference files. Use each one for its job:
- Agent Profile: my role, priorities, the people I work with and how I like answers. Use it as background whenever it's relevant.
- Communication Style Guide: how I write, including any My Voice Dial settings. Use it whenever you draft something I'll send or share. Copy its patterns, not the particulars of its examples. If one of its sections has its own instructions, follow them.
- Safety & Security Rules: how you must handle information, uncertainty, verification and human review. Follow them in every chat.

How to work with me:
- Don't invent facts about me, my organization or my work. If a missing detail would change your answer, ask me one question first.
- Run a file's First Setup Check only when I ask for it.
- My files are the master copies. When I ask to update one, follow that file's own VERSIONS section if it has one and give me the complete next version. Remind me to save it as my master file, keep the old file, and replace that file in this Project. Don't tell me the saved setup has been updated.

  2. Start a new chat in My Agent and send Check my file versions, with my rules. It should name your new rules version beside your and style guide, and quote the newest line of your rules' VERSION HISTORY. If it names your earlier rules, check that Sources lists your new rules and not the earlier file, as in step 2, then try another new chat.

    Check my file versions, with my rules
    Which versions of my files are you using?

Read the current Safety & Security Rules, Agent Profile and Communication Style Guide saved in this Project or Notebook, or attached to this chat. For each one, give its version number and updated date. If a number or date is missing, say so. Then quote the newest line of my rules' VERSION HISTORY. Use the files rather than an older conversation. Do not guess or change any file.

Step

Find ChatGPT's plugins, and who decides

Connectors let your agent work with tools you already use, such as your email, files or CRM. ChatGPT calls them plugins, some accounts still say apps, and you may also hear MCP server.

  1. Select Plugins in ChatGPT's sidebar. It lists plugins you could connect, such as Gmail, Google Drive and Salesforce. Look, and connect nothing in this workshop.

  2. Open one your work uses, such as your email. If its page groups what it can do into Read and Write actions, note which you'd need. Your workspace admin decides which plugins you can use and which of their actions are switched on. A Microsoft app may also need your Microsoft admin's consent, and a Google app your Google admin's approval.

If a plugin is missing, greyed out or marked Disabled by admin, that's your organization's answer, and nothing for you to fix. If it's one you'd want, add it to what you'll ask your IT team. If your organization doesn't allow plugins yet, you're not alone, and you can still get everything you need from this bootcamp.

Step

Know what a plugin can reach, and what you can limit

Read actions look at your information and help you with it. They can't send or change anything, so organizations usually allow them first. Write actions act for you: they draft, send, change or delete, a much bigger risk in email or any other plugin. To hear the difference for your role, send this card in My Agent.

Read-only vs write-back tools
Can you explain the difference between a read-only tool and a write-back tool in a connector?

Read access has risks too: an instruction to read only some emails can't guarantee your agent reads nothing else.

In ChatGPT, these are your controls:

  • Your admin's action settings, which decide which of a plugin's actions are switched on.
  • Always ask, if your account shows Permissions for a connected app: Settings, Plugins, the app, Connected accounts, the ⋯ menu, Settings, Permissions. ChatGPT then asks before it acts.
  • Disconnect. ChatGPT can use a connected app without being asked, and OpenAI describes no switch for one chat, so disconnect it when you're done: Settings, Plugins (or Apps), the app, Connected accounts (or Connection), the ⋯ menu, Disconnect.

Disconnecting doesn't delete chats or memories that already hold what the app read.

Step

Learn plugins from ChatGPT's own page

Plugins change fast, so anything a video shows will date quickly. OpenAI's own public documentation is your best source, and your agent can turn it into what matters for your role.

  1. Open OpenAI's Plugins page and select Copy Page at the top right. That copies the whole page as plain text your agent reads well. If you don't see it, select all of the page's text and copy that.

  2. Start a new chat in My Agent, paste the card, replace its bracketed line with the page you copied, and send it. You can add a limit to the question, as the video does: in simple terms, two paragraphs at most, only what's relevant to me.

    Explain this page in a way that matters to me
    Can you explain this in a way that matters to me?

[paste the page you copied here]

  3. Before you act on anything in the answer, find it on the page. Come back in a month and run it again, because the page will have changed.

Step

Give your agent your meetings

Here's something to use right away. A meeting transcript gives your agent the whole conversation, including the side decision nobody wrote down, and your agent can turn it into decisions and action items.

  1. Turn on transcription in your meeting app, such as Zoom or Teams, only where your organization's rules allow it, and tell everyone in the meeting it's being transcribed. If it isn't allowed, your own meeting notes work too, with less for your agent to go on.

  2. Your Safety & Security Rules and the data rules apply to a transcript like any other file. With a plugin for your meeting app that IT has approved, your agent may be able to fetch it. Without one, download the transcript, check it against the data rules, and attach that copy to a chat in My Agent rather than to its Sources.

  3. Ask your agent for the decisions made and the action items, with an owner for each. For the full method, work through Turn Your Meeting Transcripts Into Something Useful.

Agent Workshop - Connect Your Agent to Your Information Inside ChatGPT

Confirm your result

Confirm the outcomes below to finish this guide.

Final checks

A new chat in My Agent (or a chat with your three files attached) names the current version of your Safety & Security Rules (the one the versioning card gave you) beside your profile and style guide.

You can name the connector you'd want first, whether read-only is enough for it, and what your tool lets you block or switch off, and your IT team has said yes in writing, or you know what you'll ask them.

Provide Feedback

Thanks. That went straight to the people who write these guides.