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
In this repository5 other files it gives its agents.
- analyze-data-qualitySkill
- closed-loop-visual-feedbackSkill
- inverse-set-kondoSkill
- jobs-quote-uxSkill
- ux-heuristicsSkill
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
- Distinctive: line 13
Use this stack unless the person asks for something else.
- Distinctive: line 25
How the data flows: Keep these four steps separate:
- Distinctive: line 31
Run every rule over every row, not a sample.
- Very specific: line 42
Follow design-system/README.md.
- Distinctive: line 48
Never use em dashes. Not in chat, code comments, commit messages, page text, chart labels or documentation.
- Very specific: line 63
First-time setup: If the repo has no prod branch yet, create it from dev and push it:
- Distinctive: line 75
Commit after every change that works, with a short message saying what changed.
- Distinctive: line 78
Read the push output. If it did not say the branch moved, it did not push.
- Distinctive: line 80
Never force push, and never rewrite history that has been pushed.
≈ “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.
Files
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.mdwith the person, one section at a time, asking rather than guessing. When the plan changes, updatePLAN.mdfirst. 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 indist/. 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
devbranch is the test site andprodis the real one.
How the data flows. Keep these four steps separate:
- 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. - The data quality tests run against the row-level files in the cache, with DuckDB. Run every rule over every row, not a sample.
- The results are stored as small summary files in
data/summaries/: counts, rates, and a few example rows per rule. These are committed. - 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:
devis the default branch and where all work happens. Its site is the test copy.prodis the real, shared site. It only ever receives work that already ran ondev.
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 prodEvery change:
- Work on
dev. CRITICAL: DO NOT USE PRs. COMMIT DIRECTLY TODEV. 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. - Commit after every change that works, with a short message saying what changed.
- 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. - 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. - 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 devSecrets
API keys, tokens and passwords are secrets. Treat every one like a password.
- Locally a secret lives in
.env, which git ignores. Copy.env.exampleto.envand 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.
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
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.

