Module 5 of 10

Email & calendar agents.

Module 5 is the first time in the course an agent gets a keyhole into private, relational work — real inbox, real calendar, real people you care about. The directing muscle from Modules 3 and 4 transfers: read, categorize, draft, review. What changes is the safety rail. The operational norm here is two lines long and applies to every lesson: drafts, not sends. Least access for the task. You will create a bounded agent-accessible label, run a triage digest against it, draft three real email replies and send them yourself, generate a meeting-prep brief for a meeting that actually matters, and freeze the fifth capstone artifact of the course: inbox-and-calendar-log-v1.md.

Recipe Book status for this module

Module 5 is a mix of CORE and RECIPE — the Gmail and Google Calendar walkthroughs in Lessons 5.2, 5.3, and 5.4, plus the Cowork-tab triage-digest run (with a Claude Code CLI sidebar as Optional Advanced) and the permission-posture audit sequence in 5.5. Every recipe in this module was last verified 2026-04-17 and is on the next-review queue for 2026-07-17. If you hit a renamed OAuth scope, a moved Gmail label dialog, or a Cowork-tab screen that does not match the walkthrough, flag it — those drift reports are what keep the quarterly refresh honest. Outlook equivalents sidebars ship inline with the primary recipe; iCloud is flagged for a future refresh.

CORE blocks (the keyhole mental model in 5.1, the six-category triage shape in 5.2, the five-check reply review in 5.3, the five-section meeting-prep brief in 5.4, the permission-posture table and revocation ritual in 5.5) are tool-agnostic and will still be true when Gmail looks different. They are the durable part of Module 5.

Lessons in this module CORE + RECIPE

Work through these in order. Each lesson ends with a project checkpoint that adds one entry to your inbox-and-calendar log draft; Lesson 5.5 freezes the log.

  1. 5.1   What an email/calendar agent actually is.

    Mostly CORE

    The keyhole mental model — why an agent looking into your inbox is categorically different from one looking into a codebase or a research topic. The two operational norms of the module (drafts, not sends and least access for the task) stated plainly, with the third course-wide norm they install: you do not send what a human has not reviewed. You create the bounded agent-access Gmail label and the dedicated Agent Access calendar. Pairs with the Personal data surface worksheet, which becomes Entry 0 of the capstone.

    Open Lesson 5.1 →

  2. 5.2   Inbox triage — reading before writing.

    CORE + RECIPE

    Triage before drafting. The six-category digest shape (reply today / reply this week / FYIs / archivable / flagged as unclear / missed by the fence) that a well-directed agent produces against your bounded label — and the four-step audit you run on that digest every time. Recipe walkthrough for the Cowork tab (the home base for triage), with the Claude Code CLI as an Optional Advanced sidebar for terminal-comfortable students who want a scriptable, repeatable run. Ends by writing Entry 1 of the capstone inbox-and-calendar log.

    Open Lesson 5.2 →

  3. 5.3   Drafting email — the drafts-not-sends norm.

    Mostly CORE

    The most important lesson in Module 5. The five-check reply review (voice, facts, relationship fit, what is missing, what is extra), the three failure modes (smooth wrong claim, off-voice reply, over-committing sentence), and the scripted discipline of opening every draft in the Drafts folder and clicking Send yourself. Pairs with the Review-before-send drill interactive activity. You draft three real replies, review them, send them yourself, and record before/after as Entry 2 of the capstone.

    Open Lesson 5.3 →

  4. 5.4   Calendar management and meeting prep.

    CORE + RECIPE

    The five-section meeting-prep brief — what this meeting is, who is in the room, what they will bring up, what I want to land, open questions — produced by an agent from the calendar event, the thread, and any attached documents. Three kinds of calendar agent work (sweep, prep, reschedule), the RSVP and reschedule exceptions that stay human, and the difference between the parts of the brief the agent drafts and the part that is yours, always. Ends by generating a brief for a real upcoming meeting as Entry 3 of the capstone.

    Open Lesson 5.4 →

  5. 5.5   Safe patterns for sensitive data (capstone freeze).

    Mostly CORE

    The three-part definition of safe — bounded, reviewable, revocable — and the cloud-vs-local decision rule. Prompt injection stated plainly, with the four practical defenses. The permission-posture table (what / who / why / when reviewed) and the three-habit revocation ritual (revoke one-off, reduce recurring, re-audit monthly). Freeze /capstone/inbox-and-calendar-log-v1.md — the fifth capstone artifact of the course — with both safety norms pasted verbatim at the bottom.

    Open Lesson 5.5 →

Module wrap-up

End-of-module check.

Ten questions across 15 points — six multiple choice, two short-answer, two applied. Closed-book for the recall sections; open-workstation for the applied pair, where you draft a reply under time pressure against an agent’s first-pass draft and run a five-row permission-posture review end to end. Passing bar: 11.5 / 15 with full credit on at least one applied question. The parent scoring summary makes it straightforward to document as a module assessment on a transcript.

Open the end-of-module check →

Resources for this module RECIPE

Five printable companions and one interactive activity run alongside the lessons. Each printable is a print-to-PDF page that prints cleanly on letter paper. Use them in the order their lessons reference them.

  • Personal data surface — printable worksheet for Lesson 5.1. Four sections — what the agent will see, what it won’t see, what a wrong confident action would cost, who reads the output — plus a scope-and-revocation checklist for the bounded label and calendar you create in the lesson. Produces personal-data-surface-v1.md, Entry 0 of the capstone log.
  • Triage digest template — printable for Lesson 5.2. The six-category digest shape the agent is directed to produce, plus the four-step audit (count, placement, invention, missing context) you run against the digest before you trust any of it. Produces Entry 1 of the capstone.
  • Review-before-send drill — interactive HTML activity for Lesson 5.3. Four staged agent drafts walked through the five-check review one at a time, with reveal-on-click scoring that names which of the three failure modes showed up in each draft. The rep that installs the muscle.
  • Reply-draft checklist — printable for Lesson 5.3. The five-check review in a form you can tape next to your keyboard. Five checks, three failure modes, one line about what you changed. Used on every drafted reply before you send it.
  • Meeting-prep brief template — printable for Lesson 5.4. The five-section brief with Sections 1–4 tagged Agent-drafted and Section 5 (what I want to land) tagged Yours, always. Four-check brief audit, post-meeting reflection line. Produces Entry 3 of the capstone.
  • Permission-posture audit — printable for Lesson 5.5. The four-column table (what / who / why / when reviewed), the standing question (if I dropped dead or handed my laptop to a stranger...), the three-habit revocation ritual, and a monthly-cadence closeout. Produces Entry 4 of the capstone and the habit that carries forward.

What you should have when this module is done CORE

By the time you close out Module 5, you should be able to point to six concrete things in your accounts and in your capstone folder:

  • A bounded agent-access Gmail label (or Outlook folder) that contains only the messages you have deliberately routed into the agent’s surface — and an Agent Access calendar that is the only calendar the agent can read or write.
  • A Personal Data Surface worksheet saved as personal-data-surface-v1.md, with screenshots of the scoped label and scoped calendar, and a one-paragraph honest answer to “what would a wrong, confident action here cost?”
  • A saved triage digest from Lesson 5.2 — the six categories filled in against your real bounded label, with the four-step audit performed and initialed.
  • Your reply-draft checklist printed and taped where you can see it from your keyboard. If it is not physically visible to you, the five-check habit has not landed yet.
  • A frozen /capstone/inbox-and-calendar-log-v1.md containing, in order: a one-paragraph module summary, four entries (triage digest, three drafted-and-sent replies with diff notes, meeting-prep brief, permission-posture audit) each with metadata and a short reflection, and both safety norms (drafts, not sends and least access for the task) pasted verbatim at the bottom.
  • The completed end-of-module check in your portfolio, scored at 11.5 / 15 or better with full credit on at least one applied question.

If any of these is missing, go back to the checkpoint that produced it and finish before moving on. Module 6 assumes the permission posture of Module 5 is already habitual — the scheduled and long-running agents it introduces are the same discipline on work that runs without you in the room.

Coming next

Module 6 — Automation & scheduled tasks.

Work that runs without you. Module 6 picks up the directing muscle and extends it to agents that fire on a schedule — daily summaries, weekly reports, monitoring and alerts — and to the reliability, cost, and observability discipline those agents need. The safety norm extends: drafts, not sends plus least access for the task plus a new one — audience = only you.

Open Module 6 →

Module 6 opens when the Module 5 portfolio is complete — the six items above, including the end-of-module check.