Skip to content

CLAUDE.md

Latest commit

 

History

History
59 lines (47 loc) · 2.8 KB

CLAUDE.md

File metadata and controls

59 lines (47 loc) · 2.8 KB
Raw

CLAUDE.md: My Data Project

Project memory for Claude. Read this at the start of every session and follow it.

Project overview

  • Name: My Data Project
  • Type: Data / AI / automation
  • What it is: TODO: one or two sentences: what it is and who it's for.

Goals

  • TODO: your top 1-3 goals, one per line (e.g. ship a working v1 this month)

Tech stack

  • Python 3.12, pandas, Jupyter notebooks (list key libraries and versions)
  • Stick to this stack and these versions. If something here looks outdated, say so instead of silently switching.

Commands

  • Create env: python -m venv .venv
  • Install: pip install -r requirements.txt
  • Run: python main.py
  • Test: pytest

How to work with me

  • Write clean, readable code with short comments where intent is not obvious. Summarize changes briefly.
  • Before a big change, outline the plan in a few bullets and wait for my OK.
  • If you are unsure about an API, version, or file, say so and check instead of guessing.

Always

  • Keep raw data read-only; write outputs to a separate folder
  • Pin dependency versions in requirements.txt
  • Print shapes and sample rows when transforming data so results can be checked
  • Write small functions that can be tested

Never

  • Do not commit datasets, credentials, or personal data
  • Do not silently drop rows or fill missing values without saying so
  • Do not invent column names; inspect the data first

Security rules

  • Never write secrets (API keys, tokens, passwords) into code, docs, logs, or commits. Use environment variables and keep .env in .gitignore.
  • Ask before running destructive commands (deleting files, git reset --hard, force-push, dropping data).
  • Ask before adding new dependencies, services, or paid APIs.
  • Do not log or commit personal data.

Memory and backup

  • Project knowledge lives in docs/brain/ (architecture, glossary, gotchas). Read it before big changes and keep it current.
  • Record important decisions in docs/brain/decisions.md as: date, decision, why, alternatives considered.
  • At the end of each session, update docs/brain/NOW.md with: what was done, what is next, open questions. Read it at the start of every session.
  • When I correct the same mistake twice, propose a new rule for this file.
  • Commit after each working step with a clear message so any change can be rolled back.
  • Never push to a remote or deploy without my explicit OK.

Definition of done

  • It builds/runs without errors or new warnings.
  • The change does what was asked and nothing more.
  • Docs and this file are updated if behavior or rules changed.