agentleFS
dkadtsdkadts/landing-kit-building/CLAUDE.md

landing-kit-building

CLAUDE.md2 starsChanged 3 months agoMITView on GitHub

A universal system for building landing pages. Vanilla HTML/CSS/JS + Vercel. Primary CTA is a request form (no payment integrations). Standalone tools (call anytime): - /typography — re-apply widow/orphan rules to any saved HTML. Hard enforcement: - /build refuses to work until DESIGN-LOCK.md is fully LOCKED and brief/animation.md exists. - /tokens refuses to work until /design-lock is complete. - /build refuses to save a section that omits an approved-manifest item. - /deploy refuses to run if /audit verdict is not…

The file

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

6 rules. About 0 are common, 1 is distinctive, and 5 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 5

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    DESIGN-LOCK.md is frozen: If a field is marked LOCKED — do not change it.

    line 10
  2. Very specific:

    No hardcode: All colors, sizes, paddings, fonts, easing, duration — from site/css/tokens.css.

    line 12
  3. Very specific:

    User owns commits: Do not run git add, commit, push, reset — the user versions manually.

    line 16
  4. Very specific:

    Do not read site//*.{html,css,js} without explicit need.

    line 82
  5. Very specific:

    Read DESIGN-LOCK.md and site/css/tokens.css — sources of truth.

    line 86
  6. Distinctive:

    Only now — read the specific file you need to edit.

    line 88

≈ “Common” is an estimate: at least 1 in 5 of the 27,641 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

Landing Kit — Protocol for AI Agents

A universal system for building landing pages. Vanilla HTML/CSS/JS + Vercel. Primary CTA is a request form (no payment integrations).

7 rules (default for all sessions)

  1. Strict isolation. Edit only the files the user named explicitly or that clearly belong to the current task. Do not "improve" adjacent code.
  2. DESIGN-LOCK.md is frozen. If a field is marked LOCKED — do not change it. Change only through explicit user override with an entry in the decision log.
  3. No hardcode. All colors, sizes, paddings, fonts, easing, duration — from site/css/tokens.css. If a value is missing — first add it to tokens (with user approval), then use it. Hardcoded px/#hex/ms in CSS or JS is blocked.
  4. Algorithmic approach. Paddings and sizes — via Flexbox / Grid / calc() / vw. No "by eye".
  5. User owns commits. Do not run git add, commit, push, reset — the user versions manually.
  6. Typography discipline (English). No hanging short words at line end (articles, prepositions, conjunctions, auxiliaries, short pronouns, dashes) — tie them to the following word with  . No orphans — the last line of a heading or paragraph must have ≥ 2 words. Print-typography rules apply. Enforcement via /typography, called automatically by /build and checked by /audit. English-only; do not run on other languages.
  7. References priority. Every element approved during /refs, /design-lock, /animation, or /icons becomes a required entry in references/approved-manifest.md. /build will not save a section that omits a relevant manifest item. /audit fails on any unchecked item. Silent drops are the failure mode this rule exists to prevent — always ask the user to remove an item explicitly with a decision-log entry.

Workflow for a new landing (mandatory order)

1.  /brief          Survey + brief. First question: desktop-first or mobile-first
2.  /inbox          Sort client materials from inbox/
3.  /copy-analyze   Deep text analysis → brief/copy-analysis.md
4.  /refs           References parsed by aspect + approved-manifest.md
5.  /design-lock    FREEZE DESIGN-LOCK.md (spacing, type, color, motion, ...)
6.  /blocks         Config for interactive blocks (form, modal, chat, ...)
7.  /animation      Motion brief → brief/animation.md (appends to manifest)
8.  /tokens         Generate site/css/tokens.css from DESIGN-LOCK
9.  /media          Media pipeline (WebP/AVIF, resize, posters)
10. /icons          Hybrid icons: libraries + AI fallback (appends to manifest)
11. /build          Section assembly (guard against slop + hardcode + manifest coverage + typography pass)
12. /mobile         Mobile adaptation with hardened rules + safety check
13. /audit          Anti-slop + perf + a11y + typography + references compliance
14. /deploy         Vercel + optional GCP Secret Manager for sensitive keys

Standalone tools (call anytime):

  • /typography — re-apply widow/orphan rules to any saved HTML.

Hard enforcement:

  • /build refuses to work until DESIGN-LOCK.md is fully LOCKED and brief/animation.md exists.
  • /tokens refuses to work until /design-lock is complete.
  • /build refuses to save a section that omits an approved-manifest item.
  • /deploy refuses to run if /audit verdict is not OK.
  • /audit refuses to pass if site/_mobile-check-report.md verdict (from /mobile Step 7) is not OK.

Anti-AI-slop checklist (auto-enforced by /audit)

  • ❌ Emoji in production HTML (unless explicitly part of the brand)
  • ❌ More than one icon library
  • ❌ Gradients outside DESIGN-LOCK
  • ❌ More than one radius across the site
  • ❌ More than 2 font-weights
  • ❌ Center-align on more than 50% of blocks
  • ❌ Decorative shadows without function
  • ❌ More than 5 section types (repeat, do not invent)
  • ❌ Hardcoded values that belong in tokens
  • ❌ Rounded "pill" buttons if pill is not in DESIGN-LOCK
  • ❌ Line-ending on a short word (English typography rule)
  • ❌ Approved manifest items missing from the build

Sources of truth

WhatWhere
What we're buildingbrief/*.md
How it should look/workreferences/**/* + brief/copy-analysis.md
Approved required elementsreferences/approved-manifest.md
Contentinbox/**/*
Design constitutionDESIGN-LOCK.md (frozen)
Design variablessite/css/tokens.css (generated from DESIGN-LOCK)
Motion specbrief/animation.md
Iconssite/media/icons/
Decisions logDESIGN.md

Start of session

Do not read site/**/*.{html,css,js} without explicit need. Instead:

  1. git log --oneline -20 — recent changes.
  2. git status — work in progress.
  3. Read DESIGN-LOCK.md and site/css/tokens.css — sources of truth.
  4. For the task — git log --grep="<scope>" --oneline.
  5. Only now — read the specific file you need to edit.

Commit format (user commits themselves)

<type>(<scope>): <imperative> — type ∈ {fix, feat, perf, chore, refactor, docs} Examples: fix(hero): align headline to 100px margin, feat(form): add phone field validation

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.