agentleFS
dkadtsdkadts/landing-kit-building/.claude/skills/audit/SKILL.md

audit

Skill2 starsChanged 3 months agoMITView on GitHub

Pre-ship checklist. Anti-AI-slop, hardcode, perf, a11y, SEO, security, contract, mobile, typography, and references-compliance checks. Report pass/fail. Blocks /deploy on any blocking failure.

193 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use delight, teach-impeccable and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • HK-hubSkill

    delight

    Add moments of joy, personality, and unexpected touches that make interfaces memorable and enjoyable to use. Elevates functional to delightful. Use when the user asks to add polish, personality, animations, micro-interactions, delight, or make an interface feel fun or memorable.

    Also in 32 projects that use this

  • HK-hubSkill

    teach-impeccable

    One-time setup that gathers design context for your project and saves it to your AI config file. Run once to establish persistent design guidelines.

    Also in 19 projects that use this

  • zwy123zwySkill

    normalize

    将用户随意输入的提示词规范化为结构清晰、便于大模型执行的标准化提示词。适用于需要整理、统一提示词格式时调用。

    Also in 19 projects that use this

  • SkillPilotCHSkill

    level-up

    Eine Automatisierung pro Woche finden, scopen, shippen. Output landet als proposed-Eintrag in cadence.md mit konkreten Implementations-Schritten.

    Also in 12 projects that use this

  • ww-w-aiSkill

    pm-discovery

    PM Agent Team - Automated product discovery, strategy, and PRD generation. Runs PM analysis workflow to produce comprehensive PRD before PDCA Plan phase.

    Also in 3 projects that use this

  • +7 more from HK-hub

The file

We split this skill into the rules it gives an agent

5 rules. 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 174 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    After /build + /mobile (site assembled).

    line 12
  2. Very specific:

    Before /deploy (required gate).

    line 13
  3. Very specific:

    Do not run /deploy if there are fails.

    line 161
  4. Distinctive:

    Do not rewrite code for the user (dev feedback loop).

    line 162
  5. Distinctive:

    Do not skip a category — always run all of A through I.

    line 163

“Common” is measured: kept by at least 6 in 10 of the 174 edited versions of this file we could read. “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
SKILL.md

/audit

Final verifier before /deploy. Categories: anti-slop (what killed prior projects), technique (perf, a11y, SEO), contract (matches DESIGN-LOCK, brief, and the approved manifest).

When to use

  • After /build + /mobile (site assembled).
  • Before /deploy (required gate).
  • Regularly during development for sanity checks.

Check categories

A. Anti-AI-slop (DESIGN-LOCK enforcement)

Walk HTML/CSS/JS. Each check = pass / fail with file/line.

  • Emoji in HTML — grep [\u{1F300}-\u{1F9FF}] in HTML. Fail if found (unless emoji is in the brand book).
  • More than one gradient — grep linear-gradient|radial-gradient|conic-gradient. Fail if > 1 unique.
  • More than one radius — grep all border-radius:. Everything must be var(--radius). Fail on hardcode.
  • More than 2 font-weights — grep font-weight:. Fail if > 2 unique values project-wide.
  • Center-align > 50% — count sections with text-align: center vs total. Warning if > 50%.
  • More than 5 section types — count unique <section class="sec sec-...">. Fail if > 5 (see DESIGN-LOCK §8).
  • Decorative icons — visually check: icons purely for decoration? Warning.
  • Bounce/spring motion — grep cubic-bezier for out-of-range values (>1 or <0). Fail if found and not allowed in DESIGN-LOCK.
  • Mixed icon libraries — > 1 SVG sprite or > 1 icon font. Fail.
  • Pill everywhere — if DESIGN-LOCK radius ≠ 999 but border-radius: 999px present — Fail.

A2. Typography (English print-quality rules)

  • Hanging short words at line end — grep the rendered HTML for space-separated short words at potential line breaks. Specifically: scan each text node for space-terminated sequences ending in words from the /typography canonical short-word list (a, an, the, and, or, but, of, to, in, on, for, with, is, are, be, it, we, you, etc.). If the space is a regular space (not &nbsp;) and the next word exists, this is a violation. Fail with line reference.
  • Dashes at line end — grep for [-–—] (space, dash, space). If the space before the dash is a regular space (not &nbsp;), this is a violation. Fail.
  • Orphans in headings and paragraphs — for each <h1>, <h2>, <h3>, <p>, <li>, <blockquote>, verify the last two words are joined by &nbsp;. Fail if a regular space separates the last two words of a text node.
  • Language contract — if <html lang="en"> is not set, and the copy is English, fail the language attribute check (also caught in D). Typography rules only apply to English.
  • no-typo opt-outs — count elements with class="no-typo". Warning if > 3 (user disabled typography on many blocks — worth reviewing).

B. Hardcode (tokens.css enforcement)

  • Hex colors in CSS outside tokens.css — grep #[0-9a-f]{3,8} in site/css/*.css except tokens.css. Fail.
  • rgb/rgba in CSS outside tokens.css and outside var(--c-overlay) — grep rgba?\(. Fail on each.
  • Hardcoded px (except 1px border) — grep \d+px except 1px solid. Fail.
  • Hardcoded ms/s — grep \d+(ms|s)\b in CSS outside tokens. Fail.
  • Hardcoded cubic-bezier in CSS/JS outside tokens.css. Fail.
  • Inline styles — grep style="..." in HTML. Fail.
  • !important — grep. Fail (except responsive.css mobile parallax kill — allowed there).

C. Perf

  • HTML size — wc -c index.html — target ≤ 100KB post-build (without inline SVG).
  • Total CSS size — sum all CSS — target ≤ 50KB.
  • Total JS size — target ≤ 50KB (excluding external analytics).
  • Media sizes — every .webp under 500KB (special cases aside); .mp4 under 3MB.
  • Preload on hero — <link rel="preload" as="image" href="/media/hero.webp"> in head. Fail if missing.
  • font-display: swap — on every @font-face. Fail.
  • loading="lazy" on every img outside hero. Warning if missing.
  • decoding="async" on every img. Warning if missing.
  • Cache headers — vercel.json contains immutable for media. Fail if missing.

D. A11y

  • Alt on every img — grep <img without alt=. Fail per instance.
  • Aria-label on textless buttons — burger, close, prev/next carousel. Fail.
  • Label on every input — not just placeholder. Fail.
  • Contrast text/background — run WCAG formula; ≥ 4.5:1 (AA). Fail.
  • Keyboard navigation — every interactive element reachable via Tab. Manual check.
  • Focus visible — CSS :focus-visible defined. Warning if missing.
  • prefers-reduced-motion — @media rule present. Fail if missing and animations exist.
  • Skip-link — <a href="#main">Skip to content</a> first in body. Warning.
  • Lang attribute — <html lang="en"> or other. Fail if missing.

E. SEO

  • Title — not "TBD", not empty, ≤ 60 chars. Fail.
  • Meta description — not empty, 120–160 chars. Fail.
  • OG tags — og:title, og:description, og:image, og:url. Fail on missing.
  • OG image — 1200×630, ≤ 300KB. Fail.
  • Twitter card — <meta name="twitter:card" content="summary_large_image">. Warning.
  • Canonical — <link rel="canonical" href="https://...">. Warning.
  • Robots — if /thank-you exists — must be noindex. Warning.
  • Sitemap.xml — exists. Warning.
  • One H1 per page — grep <h1. Fail if > 1.

F. Security

  • Vercel headers — X-Frame-Options, X-Content-Type-Options, Referrer-Policy set. Fail if missing.
  • HTTPS-only links — grep http:// in HTML. Fail per instance (except localhost, mailto:, tel:).
  • External links rel — every <a target="_blank"> has rel="noopener noreferrer". Fail.
  • No inline event handlers — grep onclick=|onload= in HTML. Fail.
  • No secrets — grep api_key|secret|password|token= in HTML/JS. Fail.
  • Secret storage documented — verify site/_deploy-log.md or brief/06-technical.md names where sensitive keys live (Vercel env / GCP Secret Manager / other). Warning if undocumented.

G. Contract (matches brief)

  • Every section from brief/copy-analysis.md → structure present? Fail on omission.
  • Every interactive block from brief/07-interactive.md [x] implemented? Fail on omission.
  • CTA button matches brief/03-goals.md? Fail.
  • Contacts in footer from brief/01-business.md → contacts? Fail.
  • Analytics IDs from brief/06-technical.md wired? Fail.
  • Privacy Policy page exists if applicable (see brief/blocks/legal-privacy.md)? Fail.

H. Mobile

  • Viewport meta present. Fail.
  • Emulate 375×667, 390×844, 414×896 — visual smoke test. Warning on any breakage.
  • Cache-bust ?v=N on every CSS/JS link. Fail.
  • Tap-target size ≥ 44px on every interactive. Fail.
  • site/_mobile-check-report.md exists and its verdict is OK (produced by /mobile Step 7). Fail if missing or FAIL verdict.

I. References compliance (approved-manifest enforcement)

  • Manifest exists — references/approved-manifest.md present. Fail if missing.
  • All items checked — every [ ] in the manifest must be [x] by the end of /build. For each unchecked item, fail with a citation showing which reference source expected it.
  • Manifest not empty (advisory) — if the manifest has zero items even though references/ had files, warning: probably a /refs skip.
  • Removed items logged — any item that was in the manifest historically but is now absent must have a corresponding decision-log entry in DESIGN.md explaining why. Fail on silent drops (compare against git history if available; otherwise warn).

Report format

site/_audit-report.md:

# Audit report — 2026-08-08 15:30

## Summary
- ✅ Passed:      56
- ⚠️ Warnings:    3
- ❌ Failed:      2

## ❌ Blocking fails (must fix before /deploy)

1. **B — Hardcoded hex** — site/css/hero.css:23 `#0A0A0A` not in tokens.css
2. **I — References compliance** — approved-manifest item unmet:
   `[ ] motif: horizontal thin rule between sections  ← ref: mood/aesop`
   Expected in: sections separator layer. Not found.

## ⚠️ Warnings (non-blocking, discuss)

1. **A — Center-align** — 3 of 5 sections centered — risk of flatness. Keep?
2. **D — Focus visible** — :focus-visible not defined in CSS
3. **E — Sitemap.xml** — missing

## ✅ Passed
- 56 checks passed. Full list in _audit-full.md.

## Verdict
❌ /deploy blocked. Fix fails and rerun /audit.

Rules of behavior

  • Do not "improve" found problems automatically. Report to user, let them decide.
  • Fail = strict. Warning = soft.
  • Always cite file/line. Without it, the user wastes time searching.
  • Deep check once per project. Quick check for frequent runs — modified sections only.

What NOT to do

  • Do not run /deploy if there are fails.
  • Do not rewrite code for the user (dev feedback loop).
  • Do not skip a category — always run all of A through I.

Next

  • Fix fails → rerun /audit.
  • Everything pass → /deploy.

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.