The instruction block is the easy 20 percent. This ships the other 80.
You can build a declarative Copilot Studio agent in an afternoon. Getting it past security review is the part that stalls. Six agents, each finished for governance, not a demo: paste-ready instructions, knowledge-source scoping, a DPO note, a labeled-assumption cost estimate, and a test that proves the refusals fire. Then the layer after the build: a register row per agent, an approval gate with one accountable name, and the written condition that retires it.
Not sure yet? Take one agent free first: the Meeting-to-Action sample, no cost.
The instruction text is what everyone already gives away.
The six instruction blocks overlap by design with the free awesome-copilot-studio-agents repo and Microsoft's own template library. That overlap is deliberate, and it means this page has to be honest about what you're actually paying for.
The instruction block
The role, the scope, the phrasing. Genuinely useful, and free. It is the 20 percent that gets an agent running in a demo.
What gets it past review
A refusal clause proven by an included test, a hand-to-DPO governance note, and a labeled-assumption cost estimate. The undocumented 80 percent.
- A refusal clause proven by an included test script
- A one-paragraph governance note for your DPO
- A labeled-assumption cost estimate under the metered layer
- A tickable governance-and-deployment checklist and runbook
- A register row per agent: who approved it, what it reaches, whose budget carries it, and what retires it
If you only want an instruction block, take one free from the sample with my blessing. This pack is for the other 80 percent.
None of them is safety-critical. That's deliberate.
Every agent runs the same anatomy: role, scope and grounding with "Allow ungrounded responses" set OFF, refusal rules, a first-turn AI-disclosure line, and escalation to a named human.
Meeting-to-Action
Turns one meeting's notes into an accountability record. Every action resolves to a named owner and a date, or it is flagged as missing and left unassigned. It declines to judge a named person.
Policy and SOP Answer
Answers questions grounded strictly on your policy and SOP knowledge sources, with escalation to a named human the moment it is asked something outside that scope.
Vendor and Procurement Intake
Screens vendor and procurement requests against a mandatory-field checklist, with a hard stop when a required field is missing rather than guessing it.
New-Hire Onboarding Buddy
Answers new-hire questions from onboarding content only, so it never improvises company policy it was not given.
Weekly Status Digest
Summarizes the week's status updates into one digest. It reports what was said, and it never characterizes or judges the people who said it.
Inbox Triage and Draft
Triages and drafts replies for review. It carries a never-auto-send rule that also appears, and is proven, in its own test script.
One zip. A 55-page guide, six blocks, four workbooks, six standards, and one-click import packages.
Not a course to watch. Paste-ready blocks and working spreadsheets, plus the license and a read-me with a real contact address. Version 1.1, August 2026: every version is a free update from your library.
Six paste-ready instruction blocks.txt
1,500 to 2,000 characters each, as plain text so they paste into the Instructions field without markdown or smart-quote corruption. Each has [BRACKETED] fill-in slots for your org names, source paths, owner, and refusal boundaries.
Import-ready agent packages.zip, all six agents
Every agent also ships as a one-click upload package, with a one-page import guide PDF covering the upload path, what a package carries across tenants and what it does not, and what to check before you publish. Pasting the block and importing the package produce the same agent; pick whichever your environment allows.
Agent operating register.xlsx, 4 tabs, 23 fields
One row per agent: identity, the three named owners plus the co-owner who keeps the agent alive when one person leaves, what it reaches and the one line justifying that scope, its run mode, its review dates, the metered cost line and the budget that carries it, and the written condition that ends its life. Three columns calculate themselves, including whether the next review is OK, DUE SOON or OVERDUE. Six rows arrive pre-filled with the worked example.
Six operating-layer standards.md, hand one over on its own
The register field standard, the naming and ownership standard with its validator regex, the approval-gate RACI (eight gates, seven roles, exactly one accountable name each), the nine-step decommissioning SOP, the refusal fence as a clause you paste into every agent with nine tests that prove it fires, and a one-page note on running all of this next to an agent control plane.
Tickable checklists workbook.xlsx, 2 tabs
A 10-item Governance and DLP checklist with a done-when condition on each item (the Purview DLP scoping trap, least-privilege scoping, the Article 50 disclosure line, the Flex Routing DPO flag), plus a 9-step deployment runbook with a target of under an afternoon per agent.
Cost estimator.xlsx, live formulas
A zero-rating toggle and a Y/N metered switch per agent, a low, base, and high band, pre-populated with a worked example and recomputes when you overwrite the inputs.
Test-script workbook.xlsx, 7 tabs
A runnable log per agent, pre-filled with each agent's acceptance and refusal cases, with Actual and Pass/Fail columns, so you can prove the refusals before you publish.
55-page typeset guide.pdf, 11 sections
Six sections walk an agent through the same seven-part structure. The last three are the operating layer: the register and how to fill it, retiring an agent, and the refusal fence. Both checklists and all six standards print as readable pages, and it closes on a License and Contact page.
Written against a dated research brief, then checked by machine and by a human.
Slop-scan errors. The prose passed a deterministic check before a human reviewed voice and numbers.
The cost estimator recomputes to this figure and $1,034/mo at base case in a real spreadsheet engine, not a claimed number, and the register's six pre-filled rows reconcile to the same base case.
The six .txt blocks are the same tested blocks from the manuscript, not a re-typed copy that could drift.
Built for the person who already runs Copilot Studio.
This is for
This is not for
Free sample first. The Deployment Kit next.
Copilot Agent Pack, Free Sample
One agent, the Meeting-to-Action instruction block, byte-identical to the paid pack. See the shape before you buy.
The M365 Copilot Deployment Kit
The full rollout system: field guide, roadmap, governance checklist, ROI template. This pack feeds it with six fresh governed agents, deeper per-agent governance, cost and refusal treatment, and the register that tracks them after launch.
Before you buy, the honest version
What exactly do I get?
Will it work in my setup?
Isn't this the same as your free repo, or Microsoft's templates?
I already have Agent 365. Do I need this?
Is this just AI-generated fluff?
Won't six metered agents blow up my Copilot bill?
Does this help with the EU AI Act and my DPO?
What is the license, and can I get a refund?
Six governed agents, ready to deploy.
One price. Instant download. Pay once, keep the zip, and get free updates as a courtesy as Copilot changes.
Get the pack · $49Not ready yet? Take the free Meeting-to-Action sample first, no cost.