Maya and Alex Case Files: Subagent Briefs, Standing Rule and Effort Table

# Section 6 Prepared Practice Files: Subagent Reviews, Standing Permissions, Away-From-Desk Checks, and Capstone Effort Comparison

**Purpose**: Authored fictional teaching materials for Section 6 (*Delegate more without creating another job to manage (Agentic Workflows)*). Use these files to inspect the examples in Lectures `6.1` through `6.6` or test the prompts in `../prompts/section-06-prompts.md`.

---

## 1. Two-Subagent Proposal Review & Coordinator Synthesis (`6.2` — Maya / Meridian Manufacturing)

### Reviewer Briefs (`meridian-subagent-briefs.md`)
- **Coordinator**: Maya explains the decision to her usual agent in Work. That agent is the coordinator. She asks it to prepare 2 reviewer briefs for her confirmation and to pass each subagent only the records listed below, so the independent review cannot read her main conversation.
- **Subagent 1: Case-For Reviewer Subagent**
  - **Supplied inputs**: Permitted Meridian 6-week RFP scope summary, team capacity record, Maya's 4 decision criteria, **plus Maya's preferred narrower phased proposal** (core setup in weeks 1–4; custom integration in week 5).
  - **Assigned job**: Build the strongest defensible case for the narrower phased proposal and test whether its timeline and resource plan hold together.
- **Subagent 2: Independent Evidence Reviewer Subagent**
  - **Supplied inputs**: The exact same Meridian RFP scope summary, team capacity record, and 4 decision criteria — **without Maya's preferred phased answer**.
  - **Assigned job**: Evaluate the raw evidence from scratch, compare alternative ways to handle the bid, and search for facts in the records that contradict the timeline or capacity assumptions.

### Coordinator Synthesis Report (Disagreement Preserved, Never Averaged)
1. **Where the reviewers agree**: Both subagents conclude that bidding on the full 6-week scope is too risky because the active client launch has unconfirmed third-party testing through day 10.
2. **Where the reviewers materially disagree**:
   - *Case-For Subagent* treats the **20 hours of senior integration specialist time in week 5** as available for Meridian's custom ERP integration.
   - *Independent Evidence Subagent* flags a footnote in the department engineering calendar showing an **internal platform upgrade** scheduled during week 5 that may claim those same 20 specialist hours.
3. **Missing evidence**: Written confirmation of whether the internal platform upgrade in week 5 can be shifted to week 6.
4. **Judgment returned to Maya**: Confirm with the VP of Engineering whether the week-5 internal platform upgrade is movable before submitting the phased bid.

---

## 2. Scoped Standing Permission Rule (`6.3` — Alex / Northstar Reminders)

- **Before the rule: action receipts**: Alex first reviews what the agent actually did on earlier runs, such as the calendar entry it read back after each booking. A saved record like that, showing an action really happened, is called an **action receipt**. The Northstar reminders in this course are a prepared example, so there are no real receipts behind them. In your own work, set the rule only after you have checked receipts from your own workspace.
- **Rule name**: `standing-permission-personal-review-reminder.md`
- **1. Permitted action and destination**: Create one **15-minute private preparation reminder** on **Alex's personal calendar only**, 60 minutes before any confirmed Northstar meeting.
- **2. Required evidence**: The client review date and time must be explicitly confirmed in the supplied meeting transcript or project record.
- **3. Routine reporting & readback**: After creating the reminder, read back the saved entry from Alex's personal calendar and log the receipt in Alex's weekly summary report.
- **4. Immediate-return exceptions (Book nothing; return immediately)**:
  - The review date or time in the transcript is ambiguous.
  - The 15-minute slot conflicts with an existing event on Alex's calendar (for example, Tuesday's 10 o'clock department budget call).
  - The calendar tool is unavailable.
  - The event would invite, notify, or affect any other person.
- **Withdrawal / Pause rule**: During quarter-end review weeks, Alex withdraws the standing permission so all reminder actions return to `Proposed only`.

---

## 3. Away-From-Desk Continuity Check (`6.4` — Work, Dots, and Apps)

| Execution Route | Where the Task Runs | What to Verify |
|---|---|---|
| **1. ChatGPT Work, local** | In ChatGPT on your computer. | Uses the folders and files you have approved on that machine. Nothing local follows you to a phone. |
| **2. ChatGPT Work, cloud task** | On OpenAI's own computers, so it keeps going after you leave. Check on it or give the next instruction from the web or your phone. | Reaches files on your own computer only through **local computer access**: the computer must be connected, switched on and online, and in a company workspace the person who manages ChatGPT usually has to turn the feature on. |
| **3. `dot` (some plans)** | A personal agent with its own cloud computer and its own web browser. | Can connect to one of your computers only while that computer is online with ChatGPT open. Reads cloud files only through a linked app.  |

- **Apps**: A cloud service such as Google Drive, once linked to ChatGPT, is called an **app**. To keep it apart from the ChatGPT app on your phone, this course says **linked app** (older guides call it a connector). A linked app in ChatGPT's list still needs your employer's separate policy approval before you link company documents to it.
- **Maya's trip**: If Maya shuts her laptop and leaves, her phone still shows the conversation, but nothing can read the Meridian scope PDF or the saved decision record because the laptop is offline. Maya's company has local computer access set up, so before leaving she keeps the laptop switched on and online and starts a new cloud task for the next step, because a task begun on the laptop cannot switch to the cloud partway through. Her company has local computer access set up and she has sync on. Between client visits she opens that cloud task on her phone and asks the agent to check the phased proposal against the latest capacity note. The cloud task reads the files on her connected laptop and brings back the answer. A dot connected to the same laptop could do the same check from its own cloud computer.
- **On-screen note**: As of 20 October 2026: Pro plans in the European Economic Area, the UK and Switzerland do not include dots.
- **Check your own screen**: These details change often.
- **Maya's 3-part check**: (1) Does the task run locally or in the cloud, and is her laptop connected? (2) Can that route read the Meridian proposal and its current decision record? (3) Which linked apps does her organization approve for client bid documents?
- **Fallback**: If her company does not allow client bid files to be reached from outside the office, either through a connected laptop or a linked storage app, Maya has her desktop agent prepare a **permitted, non-confidential summary** of the Meridian options and decision record inside her employer-approved mobile workspace before leaving her desk.

---

## 4. Capstone Complete-Effort Comparison & Proportionate Team Sharing (`6.5`)

### Four-Bucket Complete-Effort Comparison (`Meetings and follow-through` on Northstar)

| Effort Bucket | Before the Workflow | With the Workflow |
|---|---|---|
| **1. Initial one-time setup** | None | About 1 hour in week one (organizing `Northstar Onboarding` project records + capturing the `meeting-follow-up` skill) |
| **2. Task explanation** | About 10 min each week explaining Northstar's background in a fresh chat | **1 spoken sentence**, under a minute, supplying Friday's transcript and Monday's review goal |
| **3. Output review** | About 15 min reading through a long unsorted draft | **3 minutes** on one organized page |
| **4. Manual correction** | About 15 min rewriting messages by hand | **1 minor phrasing tweak**, about 2 minutes |
| **Capstone Verdict** | About 40 min of recurring effort | About 34 minutes saved each week, more than half an hour; the setup pays back within 2 weeks. **KEEP**, using the same keep-or-change test as the first section. |

These minutes are made up for this prepared example. Alex supplies the transcript and a 1-sentence goal. The agent drafts the blocker-focused follow-ups in the style Alex's project notes describe. The standing permission from `6.3` is still paused from quarter-end, so the agent proposes the Monday preparation reminder and waits for Alex to confirm it. Change them to your own figures when you test your workflow. A single run shows how the workflow performed on this task, not a guaranteed saving every week.

### Five-Item Proportionate Team Sharing Package
1. **Saved skill instructions** (`Meetings and follow-through` + its supporting skills).
2. **One permitted worked example** (`Northstar follow-through` page and transcript).
3. **Explicit exception return rules** (pause when review purpose is missing; hold calendar reminders when conflicts or external attendees exist).
4. **Named human review ownership** (the human sender always reviews and sends every external message).
5. **Employer-approved tool requirement** (run only inside workspace tools approved by the organization).

---

## 5. The Course Principle and First Step (`6.6`)

- **Principle**: You lead the work, and the agent prepares it.
- **Four questions that follow from it**: What result would help the next decision? What grounded context and methods does the agent need? Which exceptions or conflicts must return to me immediately? Which human judgment or conversation belongs only to me?
- **First step**: Pick one meeting this week whose follow-up you usually end up chasing. Run the meeting follow-through workflow on its transcript, and judge the result with the keep-or-change test from the first section.