Build your own automation.

I want you to build me an automation that runs on my own machine.
Read all of this before you do anything.

## The job

- What I want automated: [THE TASK, e.g. "every weekday morning, tell me which
  invoices are more than 30 days overdue"]
- When it should run: [HOW OFTEN, e.g. "weekdays at 7am"]
- What it should produce: [THE OUTPUT, e.g. "a short summary I can read in
  30 seconds"]
- Where the output should go: [WHERE, e.g. "a file in this folder" or "an email
  to me"]
- I am not a developer: [TRUE OR FALSE]. If true, explain everything in plain
  English and never assume I will read the code.

## What it is allowed to touch

- It may READ: [WHAT IT READS, e.g. "the CSV I will put in this folder" or "my
  invoices inbox"]
- It may WRITE to: [WHERE IT WRITES, e.g. "a file in this folder, and nothing
  else"]
- It must NEVER send, post, reply, delete or change anything anywhere, unless I
  explicitly say otherwise later. Read only until I say so.
- Keep any keys or passwords in a local file in this folder, never in the script
  itself, and add that file to .gitignore.

## Step 1: plan before you build

Tell me, in plain English and before writing any code:
1. What you are going to build and roughly how it will work.
2. Exactly what it will read and what it will write.
3. Anything you need from me (a login, a key, a file, a decision).
4. Anything about the job I described that will not work, or that I have
   probably underestimated. Be blunt about this.

Wait for me to say go.

## Step 2: build it

Build it inside this folder, nothing outside it. Keep it as simple as the job
allows. Write a short README in the same folder that says what it does, how to
run it by hand, and how to switch it off. Write it for someone who has never
seen the code.

## Step 3: dry run it in front of me

Run it once now, in a mode where it changes nothing, and show me:

  DRY RUN
  would write    [the file it would create]
  would send     nothing
  would delete   nothing

Then tell me what it WOULD have done if this were live. If you cannot tell me
what it is about to do, it is not ready to do it.

## Step 4: schedule it

Once I have said the dry run looks right, schedule it on this machine to run
[HOW OFTEN, same as above]. Tell me in plain English how to check it is
scheduled, how to change the time, and how to stop it.

## Step 5: make it visible

Add one small page or file I can open that shows every run: when it last ran,
whether it worked, and what it produced. An automation I cannot see is one I
will stop trusting. If it fails, it must tell me, not fail quietly.

## How to treat me while we do this

Ask me one question at a time. Do not move to the next step until I have
answered. If something is ambiguous, ask rather than guessing. If I have asked
for something that is a bad idea, say so and tell me what you would do instead.