agentleFS
ExagrowExagrow/dashboard-workshop/CLAUDE.md

dashboard-workshop

CLAUDE.md0 starsChanged 37 days agoMITView on GitHub

This repo is a participant's copy of an Exagrow workshop starter. - Plan before you build. Before writing any code, fill in PLAN.md with the person, one section at a time, asking rather than guessing. When the plan changes, update PLAN.md first. Check the work against its success criteria before calling anything done. Use this stack unless the person asks for something else. It is the one the workshop's reference dashboard uses, so the room can help each other. -…

  • Reads credentials
  • Commits and pushes

The file

We split this CLAUDE.md into the rules it gives an agent

9 rules. About 0 are common, 7 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  • could be automatic: a setting could enforce it
  1. Distinctive:

    Use this stack unless the person asks for something else.

    line 13
  2. Distinctive:

    How the data flows: Keep these four steps separate:

    line 25
  3. Distinctive:

    Run every rule over every row, not a sample.

    line 31
  4. Very specific:

    Follow design-system/README.md.

    line 42
  5. Distinctive:

    Never use em dashes. Not in chat, code comments, commit messages, page text, chart labels or documentation.

    line 48
  6. Very specific:

    First-time setup: If the repo has no prod branch yet, create it from dev and push it:

    line 63
  7. Distinctive:

    Commit after every change that works, with a short message saying what changed.

    line 75
  8. Distinctive:

    Read the push output. If it did not say the branch moved, it did not push.

    line 78
  9. Distinctive:

    Never force push, and never rewrite history that has been pushed.

    line 80

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

Workshop dashboard: instructions for Claude

This repo is a participant's copy of an Exagrow workshop starter.

How to work with this person

  • Plan before you build. Before writing any code, fill in PLAN.md with the person, one section at a time, asking rather than guessing. When the plan changes, update PLAN.md first. Check the work against its success criteria before calling anything done.

How this dashboard is built

Use this stack unless the person asks for something else. It is the one the workshop's reference dashboard uses, so the room can help each other.

  • Observable Framework builds the site. Pages are Markdown files in src/, and the result is a plain static site in dist/. Charts use Observable Plot; filters use Observable Inputs. Both come with Framework.
  • Python with DuckDB does the data work, run through uv. Python is for the data pipeline only; the site itself is static.
  • Node.js 20 or newer is needed. If Node or uv is missing, install it with the person.
  • GitHub holds the code. Netlify hosts it and rebuilds on every push: the dev branch is the test site and prod is the real one.

How the data flows. Keep these four steps separate:

  1. Raw data lands in data/raw/, the local cache. It comes from the flash drive or the shared folder, or is downloaded once. Git ignores it; never commit it, and never download a file that is already there.
  2. The data quality tests run against the row-level files in the cache, with DuckDB. Run every rule over every row, not a sample.
  3. The results are stored as small summary files in data/summaries/: counts, rates, and a few example rows per rule. These are committed.
  4. Every page reads the summaries, through a Framework data loader in src/data/. No page ever reads the raw rows.

Because the raw files stay in the cache, changing a test means running it again, not downloading again.

Look and feel

Follow design-system/README.md. Before styling anything, ask once whether the person has their company's brand guide, colors, fonts, or logo to use in place of the defaults.

Writing style: no em dashes

Never use em dashes. Not in chat, code comments, commit messages, page text, chart labels or documentation. An en dash or a doubled hyphen is not a substitute. Rewrite the sentence with a full stop, colon, semicolon, comma or parentheses instead. If you find an em dash in a file you touch, fix it in the same edit.

Why: language models reach for em dashes far more often than people do, so a page full of them reads as machine written. Ordinary punctuation reads more professionally.

Branches: dev and prod

This project has two branches, and each one is a live website:

  • dev is the default branch and where all work happens. Its site is the test copy.
  • prod is the real, shared site. It only ever receives work that already ran on dev.

First-time setup. If the repo has no prod branch yet, create it from dev and push it:

git checkout dev
git branch prod dev
git push -u origin dev
git push -u origin prod

Every change:

  1. Work on dev. CRITICAL: DO NOT USE PRs. COMMIT DIRECTLY TO DEV. This repo literally only has one person working in it and they are approving what you do by reviewing it as you go. Don't make extra work for them to do PR theater for no one.
  2. Commit after every change that works, with a short message saying what changed.
  3. Push after every commit, immediately: git push origin dev. The push is the backup. Work that exists only on this laptop can be lost; work on GitHub cannot.
  4. Read the push output. If it did not say the branch moved, it did not push; fix it before doing anything else. Never use git push -q, which hides failures.
  5. Never force push, and never rewrite history that has been pushed.

Promoting to prod happens only when the person says so, in so many words ("ship it to prod", "promote", "publish the real site"). Then:

git checkout prod
git merge --ff-only dev
git push origin prod
git checkout dev

Secrets

API keys, tokens and passwords are secrets. Treat every one like a password.

  • Locally a secret lives in .env, which git ignores. Copy .env.example to .env and have the person paste the value in themselves. Do not ask them to paste it into chat, and never print it, log it, or echo it back.
  • On a hosting service it lives in that service's environment variable settings, set by the person in its dashboard.
  • Never in code, never in git.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.