Documenting Your Pipeline

---

Main quest — Write a README

Write your README while the code is still fresh — before you close the terminal.

Task: Write a README.md for your Day 3 pipeline.

This is the same README.md you’ve been building since Day 1 — keep adding to it, don’t start a new file.

Add a pipeline writeup to your README.md. Here’s a skeleton — fill in each section yourself:

# Day 3 Pipeline — SEC Form 3 Extraction

**Author:**
**Date:**

## What this does

<!-- one or two sentences: what does this pipeline extract, from what, using what? -->

## How to run

### Prerequisites
<!-- what someone needs before running: cluster access, API key, venv, ... -->

### Steps
<!-- the commands to submit the job, monitor it, and check the output -->

## Outputs

<!-- which files land where, and what's in them -->

## Data

<!-- the input data and where it comes from -->

🟢 Green sticky = I’m done and ready    🔴 Red sticky = I need help

README written and covers what the script does, how to run it, and where output lands

---

Side quests

Finished early? Try this one.

Side quest — Have Claude Stress-Test Your README

Ask Claude Code to read your README as if it were a labmate seeing this pipeline for the first time, and to flag anything that would stop them from rerunning it without asking you a question. Fix at least one thing it flags.

I had Claude critique my README as a first-time reader and fixed at least one issue it flagged

---

Go further — let Claude keep the record. You don’t have to write all of this by hand. You can have Claude draft the README straight from the code and then review it, or go a step further and keep a research log — a running record of what you ran, when, and where the results landed — that Claude updates each time you launch a job (see the claude -p batch-mode idea in Writing a Slurm Job with Claude). Documentation only earns its keep if it helps you later, so keep whatever form is most useful for your research and drop the rest.