The Curriculum / Reader / 07 — Automation
LEVEL 1 · ESSENTIALS · INDIVIDUAL TRACK

07 — Automation

This page compiles 3 files from the repository, verbatim, in reading order. The living version: this folder on GitHub.

level-1-essentials/individual/07-automation/README.md

07 — Automation

Automation is where AI moves from "helpful tool" to "leverage." Start with three specific automations.

Files in this folder

  1. first-three-automations.md — the exact three to build first
  2. local-tools-setup.md — Cursor, Claude Code, Raycast, ffmpeg, clipboard manager

The principle

Don't automate everything. Automate the three things you do most that are boring.

Definition of done

level-1-essentials/individual/07-automation/first-three-automations.md

The First Three Automations

Build in this order. Each takes 30–60 minutes.

Automation 1 — Daily brief

What it does: Every morning at 7am, you get a summary of: - Calendar for the day - Overnight email requiring reply - Overnight Slack DMs / mentions - Top 3 priorities from your goals doc - Weather + one thing to know from the news

Build it with: - Easiest: Perplexity Computer → Schedule a cron task at 7am America/Chicago, connect Gmail + Google Calendar + Slack - Alternative: Claude Code with a scheduled shell script - Alternative: Zapier / Make with a scheduled trigger + email step

Prompt for the scheduled task:

Every weekday at 7am America/Chicago:

1. Fetch today's calendar events (Google Calendar)
2. Fetch overnight unread emails (Gmail, marked important)
3. Fetch Slack DMs and mentions since 6pm yesterday
4. Read my goals doc at /path/to/goals.md
5. Summarize:

## Today
- Calendar: bullet each event with time + location + prep note
- Priorities (from goals): top 3
- Weather: 1 line
- News: 1 headline that matters

## Overnight
- Emails needing reply (max 5): sender + one-line + urgency
- Slack: mentions + DMs (max 10)

Email the summary to me. Total under 250 words.

Automation 2 — End-of-week review

What it does: Every Friday at 4pm, generates your weekly review using the prompt from 04-prompt-library/19-weekly-review.md, pulling from Calendar, Slack, GitHub, and your notes.

Build it with: Perplexity Computer scheduled task

Prompt:

Every Friday at 4pm America/Chicago:

1. Pull my calendar events from Mon–Fri this week
2. Pull my commits and PRs from GitHub this week
3. Pull the Slack messages I sent this week (mine only)
4. Pull notes I've added this week to [notes location]

Then run the weekly-review prompt from my prompt library:
[paste the prompt from 04-prompt-library/19-weekly-review.md]

Send the output to me by email and Slack DM.

Automation 3 — Inbox triage

What it does: Every 2 hours during working hours, scans unread email and: - Labels: Reply-today, Reply-this-week, FYI-only, Newsletter, Trash - Drafts replies for the Reply-today items

Build it with: Zapier + OpenAI, or Perplexity Computer with Gmail connector

Rules to give it:

For each unread email:

1. Classify:
   - Reply-today: needs response within 24h, from a real person, not a bulk sender
   - Reply-this-week: from a real person, not urgent
   - FYI-only: informational, no action needed
   - Newsletter: bulk sender, marketing, updates
   - Trash: spam-like, expired promo, notifications with no info value

2. For Reply-today items, draft a reply in my voice using the email-reply-drafter prompt from my library.

3. Never send. Only draft. Save drafts to Gmail's drafts folder.

4. Slack DM me a summary: "3 replies ready to review, 5 FYIs, 12 newsletters archived."

Progression: what to automate next

Once these three are running smoothly, look for the next three:

Anti-patterns

Do NOT automate these on day one: - Anything that sends messages to other people (too risky) - Financial transactions - Anything that modifies external systems without a review step - Anything with a false-positive cost you can't tolerate

Rule: Automations should default to drafting and summarizing, not sending or acting.

level-1-essentials/individual/07-automation/local-tools-setup.md

Local Tools Setup

The five tools every AI power-user should have installed locally.

1. A great code editor with AI baked in

Choose one:

Cursor

Zed with AI

VS Code + Continue + GitHub Copilot

2. An agentic coding tool

Claude Code (recommended)

## Voice [inherit master prompt voice]

## Codebase conventions - Framework: [Next.js / Node / whatever] - Tests: [Vitest / Jest / etc.] - Formatting: [Prettier config]

## Commands - Build: npm run build - Test: npm test - Lint: npm run lint

## Never do - Delete files without confirmation - Modify env files - Commit to main ```

3. A launcher with AI extensions

Raycast (macOS)

Alfred + AI powerpacks

PowerToys (Windows)

4. A clipboard manager

Essential when prompt-engineering — you copy and paste dozens of times per day.

Configure: 30-day history, exclude password fields.

5. Media tools

ffmpeg

# macOS
brew install ffmpeg

# Windows
winget install ffmpeg

# Linux
sudo apt install ffmpeg

Used for: converting audio/video for Whisper, extracting frames, compressing outputs.

yt-dlp

brew install yt-dlp

Used for: downloading videos to feed to transcription / analysis.

mat2 (metadata scrubber)

brew install mat2

Used for: stripping metadata from files before uploading (privacy hygiene).

6. Text expander

Snippets that expand into common prompts.

Suggested snippets: - ;;aboutme → your compact about-me block - ;;style → link or paste of your style guide - ;;reviewprompt → the code-review prompt - ;;summarize → the meeting-summarizer prompt

7. Ollama (optional but recommended)

Run local models on your machine for offline / sensitive work.

# Install
brew install ollama

# Pull a model
ollama pull llama3.3:70b
ollama pull qwen3:32b
ollama pull deepseek-r1

# Run
ollama run llama3.3

Best used for: - Sensitive data that can't leave your machine - Learning how models work - Offline scenarios

8. Environment variables setup

Create ~/.env.local (never commit) with your API keys from individual/02-accounts/api-keys-setup.md.

Add to your shell profile (~/.zshrc or ~/.bashrc):

if [ -f ~/.env.local ]; then
  set -a
  source ~/.env.local
  set +a
fi

Now every terminal session has your keys available as env vars.

Verification

After setup, run:

ffmpeg -version
yt-dlp --version
ollama --version
claude-code --version   # if installed
echo $OPENAI_API_KEY | cut -c1-10   # should show 'sk-proj-' or 'sk-'

If any fails, fix before moving on.

← 06 — Memory Hygiene Jargon Glossary →