Open your dashboard
Where can I see how my whole organization is doing, in one place?
- Source
- Internal's lesson, src/week4/internal/lessons/anw-w4-l09.md (Internal 60d90336, 2026-10-19 19:39), read by md_week4.py.
- Copy
- Internal's Markdown as published to the student shelf 10-12 (Internal 101e2662a = student 8275f21b; PR #215, receipt #218), vendored byte for byte by vendor_week4.py from the student-edition overrides; the published student tree is the same files without the "Staff only" blocks (checked on import). The drawings are the inline SVG figures in the lesson files, used as written.
- Prompts
- Word for word from the lesson file Internal ships (course/workforce/week4/). Not run end to end from Learn.
- Checked in the app
- Not checked in the app from Learn.
- Open to students
- Yes.
- Other view
- one part at a time (one part at a time is the lesson; one scroll is kept to compare)
Staff review only: full controlling prompt. Students paste only the short request below.
Use the native aibl-the-professor agent to guide the Week 4 lesson Open your dashboard. Read course/workforce/week4/lessons/anw-w4-l09.md completely, its details page course/workforce/week4/reference/anw-w4-l09-details.md when asked for more, and the native aibl-course-guide and aibl-lesson-map methods for this app. You teach and read; you never open, start or change the dashboard, and never act on anything it shows for the student. The student talks to the Chief of Staff in its own thread. Do not call, spawn or route to any agent. Do not edit any file.
Use these words exactly. The dashboard is a page of the student's own records, running on this computer only. In Claude Code the student acts on it in one click (in Codex the dashboard shows the records, and each button shows the words to say in the chat): grade a reply (Save my grades saves and seals the grades), press a Needs-you button, or type to the Chief of Staff in its box; each runs the Chief or Evy on the student's own subscription. Each button also shows the words to tell the Chief of Staff (or Evy) in the app's chat, such as "Grade the cold-food reply with me", as a backup. Nothing on the page sends, posts, deletes or pays. The dashboard ships in the student's workbench at workforce/dashboard-live/; the student opens it from the app's Preview button, "My Workforce dashboard" (or asks the Chief of Staff how); there is no terminal step, and the dashboard refuses to start from any agent run. Its tabs are Today (the morning brief and how each team did), Inbox (the review page from the email team's inbox checks in Week 3), Evals (each team's grades for each agent and for the whole job, how often the student and the Evaluation Analyst (Evy) agree, how many cases the student still checks by hand, open disagreements, both shelves with where each card came from, and the change record), Org, Knowledge, Projects and Settings. Every number on it is read from the student's records, never typed by an agent. Evals also shows each team's cost if on the API (the total for the jobs shown) and, under "What each job would cost on the API", each job's cost, its tokens and how the cost was worked out. On a Claude or ChatGPT subscription the student didn't pay this; it's what the same work would cost on the API. A job with no cost on record says "not recorded", never $0. Name agents job title first, nickname in brackets on first use; Evy is "it". Say "your work", never "business"; refer to other lessons by title, never by number.
The Hattie B's practice inbox is the worked example on the page. Never reveal or paraphrase any answer key for the practice emails; use only what the lesson page shows.
Ask open questions only.
Part 1. If asked, explain how the dashboard opens in each app, from the page, and that the student can grade, press a button or ask the Chief of Staff right on the page.
Part 2. If asked, help the student match what Today and Inbox show to their brief and their inbox checks.
Part 3. If asked, help the student find each number on the Evals tab and say what it counts, including where to see the cost if on the API and what it means. If a number looks wrong, say which record it should come from and that the student asks the Chief.
Part 4. If asked, help the student find one card's origin on the shelves and one entry in the change record.
Otherwise, if asked, explain the step and what the student should see.
Staff only: status, sources, what the build supports. Not student-facing.
Status: first draft, 8 October 2026, lane "lessons" (dispatch 51bbe9da), from spine v2 Module 5 and DESIGN-week4-2026-10-08.md section 9 (tabs Today, Inbox, Evals, Org, Knowledge, Projects, Settings; Evals shows grades for each agent and the whole job, agreement and how much the owner still checks, open disagreements, both shelves with origin badges, the change record; read-only routes in board_server.py; Field Manual restyle; rebind with prepare_workforce_dashboard.py; lift WITHHELD_PACKAGE_PATHS and hold_week4_board()).
Bound to PR #200, then the dashboard you can act on (12 Oct 2026, review fixes). The owner opens the dashboard from the app's Preview (Run) button, "My Workforce dashboard": the course launch entry (workforce/house/.claude/launch.json, added by launch-merge.py) runs python3 workforce/dashboard-live/helper.py --workbench . --app claude --port 8791 with url http://127.0.0.1:8791. No terminal step. No agent may start it (no house allow line; every dashboard run denies it; the helper refuses to start when AIBL_DASHBOARD_RUN, CLAUDECODE or Codex's own variables are set; the app's own launch sets none of them, checked 12 Oct 2026). The key is never printed or written to disk. The owner-only folder (Application Support/AIBL, ~/.config/aibl or AppData/Roaming/AIBL) holds only the view-mode switch; the house settings and every run deny it, and Evy's guard denies naming it. It reads the workbench through workforce/runtime/board_server.py, refuses every write without the key and any request marked as Evy, runs every Claude turn with Evy's guard in its own settings file, and stops a run at its first event unless it is on the subscription. Actions are on (second Week 4 update, 12 Oct 2026). The key lives only in the helper's memory and the owner's tab: the helper opens the owner's browser at #c=<one-time code>, which the page trades once for the key (POST /api/claim). Before any run it checks the app's sign-in without a model request (claude auth status --json: claude.ai, first party, a subscription; codex login status: ChatGPT) and starts nothing otherwise. Evy's verdict on a job in the owner's blind sample is hidden on every tab (workforce/runtime/evy_hold.py). Runs from the dashboard run no course script (Python, .mjs, .sh and anything under workforce/ are denied; a script reads past the run's Read denies), and while a job is held they cannot read its evaluation, Evy's daily folders, the briefs, the Chief's judgment log, agent memory or Evy's notebooks, and have no Grep. /api/grade seals through agreement.py run by the helper itself. Runs from the dashboard are Claude Code only: with --app codex the dashboard stays in view mode, because Codex's sandbox can write the whole workbench and has no read fence the dashboard could set; every run's settings deny writes to the course's own folders (workforce/, .claude/, .codex/, .agents/, course/, .aibl-local/). View mode (every write and run route 404, each button shows only the words) is the owner's switch: a 0600 file outside the workbench, <owner folder>/<workbench id>.actions.json, {"actions": false}. It has seven tabs (Today, Inbox, Evals, Org, Knowledge, Projects, Settings). The dashboard entry still opens the read-only board.
- Screen. Part 3's second figure draws the Hattie B's Evals screen from the tab contents PR #200 lists. Swap it for a screenshot of the board running on the practice fixture (PR #200's visual check used one) when the integration render is made.
- The Hattie B's numbers below are a made-up worked example, consistent with "Compare your grades with Evy's and teach Evy".
| Figure ID | Lesson | What it shows | Caption | Alt text |
|---|---|---|---|---|
w4-l09-tabs | anw-w4-l09 Part 1 | A drawn browser window with a row of tabs (Today, Inbox, Evals, Org, Knowledge, Projects, Settings), the Evals tab selected, and three panels under it: grades for each agent and the whole job; agreement and how much you still check; shelves and change record | One page for everything your organization did, read from your own records. | Your dashboard has seven tabs: Today, Inbox, Evals, Org, Knowledge, Projects and Settings. The Evals tab shows each team's grades for each agent and for the whole job, how often you and Evy agree and how much you still check by hand, and the two shelves and the change record. |
w4-l09-evals-hattie | anw-w4-l09 Part 3 | The Evals tab for the Hattie B's email team: a team picker set to "email-team"; a grades panel with the whole job 4 pass and 1 fail and the writer's one fail; an agreement panel with 19 of 20 and 4 of 4 and "check every case"; a shelves panel with two golden examples and six golden cases; a change record line "made-up food facts: kept" | The Hattie B's Evals tab after the first fix: every number read from a record. | The Evals tab for the Hattie B's email team shows the latest rerun with four passes and one fail, the writer's one failed rule, 19 of 20 job rule grades agreeing with Evy and 4 of 4 on the ones the student failed, every case still checked by hand, two golden examples, six golden cases, and the change for made-up food facts marked kept. |
|---|
Everything you've made this week lives in files: grades, rule sheets, Evy's notebook, shelves, the change record. Today you open your dashboard: one page that reads those files and shows them together.
You act right on the page: grade a reply yourself, press a button on something that needs you, or type to your Chief of Staff in its box. Each button also shows the words to say to your Chief, if you'd rather say them in its thread. Nothing on the page sends, posts, deletes or pays. When your Chief answers from the page, it reads and writes your files and uses its team, but it doesn't run course checks or scripts there; for those it asks you to carry on in its chat thread. Saving your grades on the page seals them right away.
Where it lives and how to start it. The dashboard ships in your workbench, in the folder workforce/dashboard-live/: the page and the small program that serves it. Open your dashboard from the Preview button and pick "My Workforce dashboard", or ask your Chief how. It opens in your own browser, ready for you to act on in one click. Each button also shows what to tell your Chief.
Your dashboard, running on your own work.
One look tells you how each team is doing, without opening a single file.
The Course Companion (The Professor) guides you in its own thread. You open the dashboard from the app's Preview button; your Chief of Staff can tell you how.
You open it, find each thing you made this week, and check the numbers match what you know.
Before you start
- Evy's daily check is on. You turned it on in "Turn on Evy's daily check".
- Nothing to install. The dashboard is already in your workbench.
- In Codex, start your first message to an agent in a thread with "Ask the <agent> to do this, and show me its reply word for word:". Start every message after that with "Tell the <agent> this, word for word, and show me its reply word for word:".
Start or resume this lesson
Open your prepared workbench and call The Professor by its role name, aibl-the-professor. Paste only this.
Use the native aibl-the-professor agent to start the Week 4 lesson Open your dashboard. Read course/workforce/week4/prompts/anw-w4-l09.txt and course/workforce/week4/lessons/anw-w4-l09.md. Ask me one question at a time and wait. I will talk to my Chief of Staff in its own thread.
The full instructions your agent follows for this lesson are in course/workforce/week4/prompts/anw-w4-l09.txt.
You ask your Chief to open your dashboard. You get one page of your organization's records.
One page for everything your organization did, read from your own records.
Do this now: open it
In Claude Code: start a new thread, type @ and pick aibl-chief-of-staff. In Codex: start a new thread and begin with "Ask the Chief of Staff to do this, and show me its reply word for word:". Use this thread for the rest of the lesson. Say:
How do I open my dashboard?
What you should see
Your Chief tells you to press the Preview button and pick "My Workforce dashboard". Your dashboard opens in your own browser, at an address that starts with http://127.0.0.1. That window is the one you act in. If the app also shows the page beside the chat, that copy only reads, and its buttons show the words to say. It works on this computer only. Seven tabs across the top.
In the Claude app, Preview is the globe icon at the top right of your Code session. Click it and pick "My Workforce dashboard".
You check the first two tabs against what you already know. You get confidence the dashboard is reading your records.
- Today shows your latest brief, including "How your teams did yesterday".
- Inbox shows the review page from your email team's inbox checks: Needs you, Team, Handled and Skipped.
Do this now: compare with your brief
Open Today. Does it match the brief your Chief gave you this morning? Open Inbox. Is your latest inbox check there? If anything's missing or different, tell your Chief in the same thread.
You find your team's grades, how often you and Evy agree, and how much you still check. You get this week's work, in numbers you can trust.
On the Hattie B's practice inbox
On the Evals tab for the email team, the student saw this:
The Hattie B's Evals tab after the first fix: every number read from a record.
Every number is read from the student's records: the rerun, the counting script, the decisions. No agent typed them.
What the work would cost. The last box in "How it's going" is Cost if on the API: the total for the team's jobs shown. Open What each job would cost on the API for each job's cost, its tokens and how the cost was worked out. On a Claude or ChatGPT subscription you didn't pay this; it's what the same work would cost on the API. So when a Hattie B's catering reply shows a dollar amount, that's what the one reply would have cost if you paid for every word it read and wrote. On your subscription it used part of your usage limit instead. A job with no cost on record says "not recorded", never $0.
Do this now: find your numbers
Open Evals and pick your team. Find the grades for the whole job and for each agent, how often you and Evy agree, how many cases you still check by hand, and what the team's jobs would cost on the API. Do they match what you saw earlier this week? If a number looks wrong, ask your Chief, in the same thread, which record it comes from.
You find the cards you saved and the change you made. You get everything from this week in one place.
Do this now: find one card and one change
Still on Evals, find:
- Your golden examples and golden set. Pick one card. Does it say where it came from? Any card waiting for review is marked.
- Your change record. Find the fix you made this week. Is it marked kept or undone, with the before and after?
The other tabs, Org, Knowledge, Projects and Settings, show your agents, your library, your work in progress and your organization's settings. Open each one once, so you know where they are. In Settings you can pick the time Evy grades each day. Your Chief then helps you set your morning digest to the same time.
If you get stuck
| What you notice | What to do |
|---|---|
| The dashboard doesn't open | Ask your Chief, in its thread, what stopped it, and wait. |
| A tab is empty | It shows only what's in your records. An empty Inbox means no inbox check has run yet. |
| A team shows no grades | A team with no rule sheet is held, not graded; write its sheet to start Evy's grading. |
| A number doesn't match what you saw | Ask your Chief which record it reads. A mismatch is a fault to fix, not your mistake. |
| A job's cost says "not recorded" | Its cost wasn't recorded, or no price list covers the model it used. Ask your Chief which record it reads. |
| You want to change something you see | Press its button. If that window only reads, say the words the button shows to your Chief, in its thread. |
| You want to show someone | It runs on your computer only. Show it on your screen, or wait for your proof page later this week. |
What each tab reads, and where: more detail on this lesson.
What you'll keep
Your dashboard, running on your own work: your brief, your inbox, each team's grades, how often you and Evy agree, your shelves and your change record.
Before you finish
Self-assessed. No gate.
Words you now own
- Dashboard: one page of your organization's records, on your computer, where you grade, press a button or ask your Chief.
- Evals tab: where each team's grades, agreement, shelves and change record live together.
- Cost if on the API: what the same work would cost if you paid the API for it. On a Claude or ChatGPT subscription you didn't pay this.
Next: Make sure everything your agents do is recorded.