Example setup

An illustrative ecosystem, separate from your learning workspace.

Explore the example · Task playbook

How an agent carries out a task

Task playbook

Choose a task below to see why it matters, what starts it, the tools involved, and the steps to follow. Review the result before allowing changes to your real accounts. Bots & tasks shows who runs the work and when; this page explains how. The agent guidelines and security checklist apply to every routine.

Weekday morning plan

When it runs: Tue–Fri, 7:30 AM (or 10 minutes before the first block) · Starts when: Interactive planning; separately configured schedules require verification

Nothing from a recording goes straight onto the main list. Staging first, then a short spoken triage, then named time.

Approach. Prompt chaining + human gate·Tools. Wispr Flow, Plaud MCP, Todoist, Google Calendar, Google Drive, GitHub

  1. 1

    Step 1

    Read `agents/daily-planner/coaching.md` from GitHub Brain, the current weekly plan in GitHub Brain, and the latest shared GitHub Brain observations.

  2. 2

    Step 2

    List Plaud files from the last 24 hours. Extract actions as Todoist subtasks under **Plaud Staging Inbox**, labels `plaud` + `staging`.

  3. 3

    Step 3

    Write a dated summary of new recordings into the Plaud Staging folder in Drive.

  4. 4

    Step 4

    Read Calendar. Default start 7:30. If the first event is earlier, plan ten minutes before it.

  5. 5

    Step 5

    Talk (voice or Wispr): staging, weekly goals, free blocks. Promote, defer, or drop together.

  6. 6

    Step 6

    Schedule the actual Todoist tasks with dates, times, and durations. The native Todoist Calendar integration supplies busy entries; do not add standalone task events.

  7. 7

    Step 7

    Update the weekly plan in GitHub Brain. Append a short dated note to shared GitHub Brain observations.

Starter prompt

You are my weekday morning coach. Be warm, specific, and brief.

1. Read the current GitHub Brain AGENTS.md and agents/daily-planner/coaching.md.
2. Read the current weekly plan in GitHub Brain and the latest dated notes.
3. Pull new voice recordings from the last 24 hours. Extract actions as staging tasks — do not put them on the main list.
4. Read today’s calendar. If the first event is earlier than 7:30, start 10 minutes before it.
5. Open a short voice (or Wispr dictation) triage: staging items, weekly goals, today’s free blocks. I decide what is promoted.
6. Schedule actual Todoist tasks with dates, times, and durations; do not create duplicate Calendar events.
7. Append a short dated note so tomorrow’s run has memory.

Never send email. Never promote staging items without my say-so.

## Ideal Week alignment

Read the current private Ideal Week alongside goals and real provider commitments. Honor hard constraints, prefer the stated days, session lengths and meeting clusters, and explain departures from the ideal. Never turn illustrative numbers into preferences. Saving an ideal plan does not authorize booking or moving existing appointments.

For future agent work, follow `workflows/agent-tasks/README.md`: capture in Todoist, review in an intentional AI tasks session, select and scope before executing. Capture never starts work automatically.
  • Cadence is guidance. Use current GitHub Brain context and live provider availability; do not infer runtime configuration.
  • Schedule actual Todoist tasks with times and durations. Avoid duplicate standalone Calendar task events.