Prompt Library

Copilot Studio Agent Prompts From a One-Job Brief

15 copy-paste prompts

Define Copilot Studio agents from one-job briefs: topics, knowledge, refusal lines, and stop before external send. Official Microsoft Copilot agents only.

In short: This page contains 15 copy-paste ready prompts, organized into 5 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Hand-curated and tested by the AI Academy team.

By Louis Corneloup · Founder, Techpresso

Last updated September 15, 2026·Hand-curated & tested by the AI Academy team

One-job agent brief

3 prompts

Agent job sentence

1/15

✨ What it does

Copilot drafts a one-sentence agent job for [NAME] with [DONE] and [NOT]. If job needs two sentences, split into two agents.

Help me write a one-sentence job for Copilot Studio agent [NAME] serving [WHO]. Done means [DONE]. Out of scope [NOT]. No invented connectors.

💡

Pro tip: One job per agent is Copilot Studio best practice, not optional.

Topic list from brief

2/15

✨ What it does

Copilot lists up to 8 Studio topics with triggers and human escalation from [PASTE]. Build topics in Studio in priority order.

From brief [PASTE], list Copilot Studio topics with trigger phrases and escalation to human when [CONDITION]. Max 8 topics. Unknown integrations [VERIFY].

💡

Pro tip: Too many topics overlap and confuse routing.

Refusal and safe fallback

3/15

✨ What it does

Copilot writes refusal text with [SAFE PATH] for [FORBIDDEN] requests. Paste into agent system topic and test once.

Write refusal text for agent [NAME] when user asks [FORBIDDEN]. Offer [SAFE PATH]. 80 words. Official Copilot safety only, no jailbreak.

💡

Pro tip: Refusal without safe path sends users to random web search.

Prompts get you started. Tutorials level you up.

A growing library of 300+ hands-on AI tutorials. New tutorials added every week.

Start 7-Day Free Trial

Knowledge and sources

3 prompts

Knowledge source map

4/15

✨ What it does

Copilot maps knowledge sources [LIST] with cite rules and conflicts. Upload exactly those files in Studio before publish.

I will attach sources [LIST] to agent [NAME]. For each, when to cite, when silent, and conflict rules vs [OTHER SOURCE]. No files not in list.

💡

Pro tip: Silent sources should trigger [VERIFY] answers, not guesses.

Answer template with citations

5/15

✨ What it does

Copilot writes an answer template for [TOPIC] with citations and [VERIFY] on silence. Use as generative answers style guide in Studio.

Write answer template for topic [TOPIC]: short answer, bullet steps, cite source name, [VERIFY] if source silent. For [WHO] tone [TONE].

💡

Pro tip: Templates keep multi-topic agents sounding consistent.

Stale knowledge audit

6/15

✨ What it does

Copilot flags stale or duplicate knowledge in [LIST] and recommends refresh order. Update files in Studio, then retest top topic.

Review knowledge list [LIST] for agent [NAME]. Flag docs older than [N] months or duplicate. Recommend refresh order. No new URLs invented.

💡

Pro tip: Stale SharePoint links are the top Copilot agent failure in enterprises.

Test in Studio

3 prompts

Happy path test script

7/15

✨ What it does

Copilot writes 5 test utterances for [TOPIC] with expected shapes. Run in Studio test pane before channel publish.

Write 5 test utterances for agent [NAME] topic [TOPIC] including one with [BRACKETS] user must fill. Expected topic match and answer shape only.

💡

Pro tip: Include one messy real user utterance, not only clean triggers.

Escalation test

8/15

✨ What it does

Copilot writes 3 escalation utterances with handoff text for [CONDITION]. Verify handoff route exists in Studio before go-live.

Write 3 utterances that must escalate to human for agent [NAME] when [CONDITION]. Include expected handoff message text.

💡

Pro tip: Escalation without route frustrates users worse than no agent.

Knowledge miss test

9/15

✨ What it does

Copilot defines expected behavior when [Q] is not in [LIST]: silent source plus [FIELDS]. Tune generative answers until it matches.

Ask agent [NAME] question [Q] not covered by [LIST]. Expected behavior: say source silent, collect [FIELDS], no invented policy.

💡

Pro tip: Knowledge miss tests prevent confident wrong HR answers.

Like these prompts? There are full tutorials behind them.

Learn the workflows, not just the prompts. 300+ easy-to-follow tutorials inside AI Academy — and growing every week.

Try AI Academy Free

Channels and guardrails

3 prompts

Teams channel intro

10/15

✨ What it does

Copilot writes a 120-word Teams intro for [NAME] from [PASTE]. Pin the post when the agent joins the channel first.

Write Teams channel intro post for agent [NAME]: what it does, what it refuses, hours, human contact [CONTACT]. 120 words from brief [PASTE].

💡

Pro tip: Intro posts set expectations better than tooltips users skip.

PII guard topic

11/15

✨ What it does

Copilot drafts PII guard topic from [POLICY PASTE] with [SAFE ALTERNATIVE]. Align with admin DLP settings before enable.

Draft a PII guard topic: detect requests for [PII TYPES], refuse collection, offer [SAFE ALTERNATIVE]. Align with Microsoft Copilot DLP if [POLICY PASTE].

💡

Pro tip: PII guard topics complement DLP, they do not replace it.

External user boundary

12/15

✨ What it does

Copilot writes external boundary message without internal doc leaks. Test with external test account if Studio allows.

Agent [NAME] may serve [INTERNAL ONLY OR EXTERNAL]. Write boundary message for external users who exceed scope. No internal doc titles leaked.

💡

Pro tip: External agents need shorter knowledge packs than internal ones.

Stop before production

3 prompts

Pre-publish checklist

13/15

✨ What it does

Copilot writes a pre-publish checkbox list for [NAME]. Owner [OWNER] signs off before you flip the production toggle.

Write pre-publish checklist for Copilot Studio agent [NAME]: topics tested, knowledge dated, escalation path, refusal tested, owners [OWNER]. Checkbox format.

💡

Pro tip: Production toggle without owner is how orphan agents stay live.

Kill switch plan

14/15

✨ What it does

Copilot writes kill switch steps with fallback message and [LOG OR VERIFY]. Store the plan next to the agent runbook.

Write kill switch plan if agent [NAME] goes wrong: disable topic order, fallback message, comms template, log location [LOG OR VERIFY].

💡

Pro tip: Kill switch plan is five minutes now, five hours during incident.

Refuse auto-send externally

15/15

✨ What it does

Copilot refuses auto-email to [EXTERNAL LIST] and proposes draft-for-human-send. Keep humans on external send unless policy explicitly allows.

Act as Copilot Studio designer. I asked agent to email [EXTERNAL LIST] automatically. Refuse. Offer draft-for-human-send pattern instead.

💡

Pro tip: Auto-send agents need legal review. Default refuse in prompts.

Free tool

Prompt Optimizer

Turn a rough idea into a structured, professional AI prompt.

Try it free →

Frequently Asked Questions

No. These prompts stay inside Copilot's official product. It uses official Microsoft Copilot Studio agents only. If a setting is missing on your plan, the prompt says so.

A current Copilot account. Some features sit on paid plans. The prompts still work as checklists if a feature is off.

Every prompt tells it to use only what you paste and to tag gaps as [VERIFY]. You still check the Copilot Studio agent before you save it.

Yes to copy. Running them needs your Copilot account. Long files fit better on a paid plan.

No. It is a draft of your Copilot Studio agent. You own the facts, the files, and whatever your job or policy requires.

Prompts are the starting line. Tutorials are the finish.

A growing library of 300+ hands-on tutorials on ChatGPT, Claude, Midjourney, and 50+ AI tools. New tutorials added every week.

Start 7-Day Free Trial Explore AI for Marketers

7-day free trial. Cancel anytime.

Browse All Prompt Collections

Popular prompt collections

AI Photo Editing Prompts ChatGPT Jailbreak Prompt Best ChatGPT Prompts Funny Hinge Prompts ChatGPT Caricature Prompts