🔵 The Greenlight
🚧 Work in progress
Scenario 2 is still being built and tested. Steps, downloads, and screenshots may change before the event.
You'll build this in Microsoft Scout, with GitHub Copilot CLI running the council behind a live dashboard. Scout does the building – you won't hand-write the app.
You run the council, then turn it into something you can watch.
What you're solving
One review is not enough when different audiences need different things from the same content. A formal explainer might help a compliance officer make a careful decision and still be unusable for a store manager who needs one practical action during a busy shift.
A single reviewer pictures one average reader and misses that gap. Here you seat a council of audiences, review the same content from each point of view, and put the whole room on a live board so the disagreement is something you can see.
What you'll walk out with
| What you make | What it does |
|---|---|
| Your council | Two or more audiences and what each one needs, in THE-COUNCIL.md. |
| A review | Shows what each audience thinks, with a quote, a source, and a confidence. |
| A live dashboard | Each seat lights up with its verdict as you feed the board content. |
| A one-step run | A simple way to start the board – a command or a scheduled task. |
The solo critic stays unchanged – the one-verdict "before" you compare against.
Before you start
Download both and keep them side by side.
Open Microsoft Scout and add the Greenlight skill. The dashboard step later also uses GitHub Copilot CLI and Node – Scout installs what the app needs, but the CLI has to be signed in and working.
1 · Import the skill and load the content
- Extract
the-greenlight.zipskill to a folder. - Add
the-greenlightskill to Scout by going to Extensions and select Import - Drag the extracted skill folder into Scout:
- Start a new chat in Scout and upload data-pack.zip.
2 · Seat the council
A seat is one audience – who they are, the outcome they need from the content, and the specific bars it has to clear for them. Seating a council means putting two or more of these in THE-COUNCIL.md so the same piece gets judged from each point of view at once.
The goal isn't to list audiences – it's to seat ones that disagree. Give each a different outcome and needs that are true for them but not for everyone (a bar any reader would score the same is just "good writing," and the solo critic already covers that). That difference is what makes the room split on a piece instead of nodding along.
Start with the provided profiles – Retail and Compliance are the sharpest pair – or add custom audiences you actually write for. Ask Scout to do the seating:
"Use Greenlight to seat the Retail and Compliance audiences from the data pack onto the council."
Done when: THE-COUNCIL.md holds two or more audiences with clearly different goals.
Seat a real audience with Work IQ
Scout is grounded in your Microsoft 365 work through Work IQ – nothing to connect and no sign-in to set up, and it only ever sees what you already can (your mail, sent mail, calendar, Teams, files, and org chart). Ask it to draft an audience profile from those real signals – "Scout, build an audience card for [a role you write for] from Work IQ that mimics the audiences in THE-COUNCIL.md" – then treat it as a first draft and correct it with people who know that audience. Keep the data in your own tenant, out of shared spaces.
3 · Convene the council
Convening is the heart of the exercise: hand the seated council a piece of content and have every audience judge it against its own criteria at the same time. Each seat returns a verdict – Ship, Revise, or Reject – and has to back it with a quote from the piece, a source (a line from its profile or the style guide), and a confidence. A seat that isn't sure says so instead of guessing.
Ask Scout to convene on the executive summary (P4) – the piece chosen to prove the point. It's a polished governance write-up: exactly what a Compliance reader wants, and unusable for a Retail floor lead who needs one action in a short standing read. A single reviewer can only average that into a mushy "needs work." A council says both things at once, with receipts:
| Piece | Solo critic | Council |
|---|---|---|
| Executive summary (P4) | Needs work (one verdict) | Reject for Retail · Ship for Compliance |
That split – the same document, opposite verdicts, each evidence-backed – is the result of this step. The thinking is done here, with nothing installed; everything after is about making it visible and repeatable. (The solo critic's scores ship recorded – they're the fixed "before," so leave them as-is.)
Done when: two seats return different verdicts on P4, each with a quote behind it.
4 · Build the dashboard
Now make the disagreement visible. Ask Scout to build a local web dashboard for the council, using GitHub Copilot CLI as the backend that runs the Greenlight skill.
Describe what you want:
- a card or avatar for each seat that lights up with its verdict – green for ship, amber for revise, red for reject
- the quote and confidence behind each seat's call, on its card
- a place to drop in a document or paste a link for the council to review
- the conflicts and who's served, at a glance
- a creative theme for the council or a custom name
Scout scaffolds the app and wires it to the Copilot CLI backend. When you feed the board a piece, the CLI runs the council and the cards update.
Stuck on the prompt? Start with this
Paste this into Scout, then adjust from there:
Build me a local web dashboard for the Greenlight council. Use GitHub Copilot CLI as the backend to run the Greenlight skill: a small Node web server that shells out to
copilot, with a plain HTML/CSS/JS front-end – no build step, minimal dependencies, so it starts with one command. Show a card for each seat inTHE-COUNCIL.mdthat lights up with its verdict (green = ship, amber = revise, red = reject) and shows that seat's quote and confidence. Add a place to drop in a document or paste a link, and a summary of the conflicts and who's served. Start simple – I'll ask for more.
The real trick is to start small and layer on: get the cards lighting up on one piece first, then ask for one addition at a time (a theme, a coverage view, animations) instead of everything in a single prompt.
If the app won't start
Setups vary – Node versions, dependencies, CLI sign-in. If the dashboard won't run on your machine, keep going in the Scout conversation; the council still works there. Get the board up if you can, but don't let it block the review.
Done when: you feed the dashboard a piece and watch the seats light up with their verdicts.
5 · Make it easy to run
Turn the dashboard into something you start in one step, so a teammate can open the board without wiring it up again. Ask Scout for either:
- a single start command – install once, then one command boots the server and opens the browser, or
- a scheduled task that launches it so the board is always there.
Unsure how? Ask Scout!
Ask Scout what the right option is for your situation. One solution doesn't fit everyone's needs, much like the different audiences in this exercise. Ironic, don't you think?
Done when: you (or a teammate) can start the board in one step and drop in new content.
Go further – the bonus
Once the board runs, the bonus is adding features to it. Keep each one small and let Scout build it:
- drag a document or paste a link straight onto the board to convene without the conversation
- a coverage view – every article by every audience, at a glance
- a history, so you can watch a piece improve after a rebuild
- a button that sends the greenlit plan to a person
- animations for when the council is deliberating or different verdicts
Show it off
60–90 seconds. Show:
- [ ] The two audiences you seated and the goal each one defends
- [ ] The solo critic's one verdict on P4, then your council splitting it – each seat with a quote
- [ ] The dashboard lighting up each seat's verdict on a piece you feed it
- [ ] One conflict and the coverage, visible at a glance
- [ ] The board started in one step – a command or a schedule
- [ ] Any feature you added to the board in the bonus round
What to aim for in the demo
Drop one piece into the board and let two seats disagree in real time. That is the thing a single generic review can never show.
Stuck?
| What you're seeing | What to do |
|---|---|
| Scout ignores the skill | Start a new session – skills load at the start. |
| Both seats give the same verdict | Give them different goals and make their needs specific. |
| The dashboard won't start | Check that GitHub Copilot CLI is signed in and Node is available; meanwhile, convene in the Scout conversation. |
| The board shows nothing back | Confirm the skill is imported and the CLI can run it on its own first. |
| The council can't see the article | Give Scout the content pack, and point the board at the same files. |
🎬 Nobody nails it first try
Let Scout off the leash and it might build you forty booths. When it overshoots, rein it in and re-run – steering the agent is the build.