PageSmith Studio Get it on Agensi

Pages that look human. Convert. Get cited by AI.

PageSmith Studio is a skill for Claude Code, Cursor and Codex. It researches before it designs, proves before it claims, and passes five quality gates before it ships. Twelve modes. Five scripts. Nothing to install.

Get it on Agensi See the three demo pages

$25, one-time. Updates included. Money-back guarantee handled by Agensi.

You can spot an AI-built page in two seconds.

Soft blue on white. A purple gradient. Three cards with icons in circles. A row of numbers nobody can source. A headline that welcomes you to the future of something. Your agent did not choose any of it. It defaulted to it, because it designed from memory instead of looking at the business.

PageSmith Studio changes the order of operations. Look first. Design second. Prove before claiming. Pass the gates before saying done.

The default page for the Turno brief: purple-to-blue gradient hero, centred welcome headline, three buttons, a floating glass mock-up The same brief built with PageSmith Studio: the Turno page, dark kitchen photograph, one call to action and a three-dimensional rota board AI lookPageSmith
Same brief. Default output on the left, Turno built with PageSmith on the right. Drag to compare.

Research first. Design second. Gates before done.

The eight steps of /page build. Each one writes a file you can read, argue with and reuse.

  1. 01

    Brief

    Paste a URL or answer eight questions. The skill extracts the offer, the audience, the proof and the stack, and asks only for the gaps.

    page/brief.md
  2. 02

    Research

    Category leaders, the colours that physically exist in your business, your customers' own words, the ten questions people ask AI engines. Three tracks in parallel.

    page/research.md
  3. 03

    Three directions

    Three competing designs scored on five criteria. The winner needs a written reason to beat the other two. The first idea is the one every other agent would have had.

    page/design-directions.md
  4. 04

    Copy

    Message hierarchy, three hero candidates, every section, microcopy. A claims policy: real proof or an honest substitute, never an invention.

    page/copy.md
  5. 05

    Build

    Static HTML and CSS, Astro, or your existing Next.js. At most 150 KB of JavaScript. Full content in the HTML, because AI crawlers do not run scripts.

    dist/
  6. 06

    SEO + GEO layer

    Metadata, JSON-LD generated from the brief, answer blocks engines can lift verbatim, llms.txt, and a robots policy that lets AI crawlers in. Then a test that your CDN is not sending them 403.

    schema.jsonld · llms.txt · robots.txt
  7. 07

    Five gates

    Tells, CRO, GEO, performance and accessibility, trust. PUBLISH, REVISE or HOLD, with the top fixes first. The page ships only on PUBLISH.

    page/gate-report.md
  8. 08

    Hand-off and deploy

    Design rationale, proof used and missing, the one metric to watch. Deploy recipes for Cloudflare Pages, Vercel, Netlify and GitHub Pages, with crawler and schema checks after.

    page/PAGE-NOTES.md

Five gates. One verdict.

An adversarial reviewer that did not build the page and does not like it yet. HOLD on any critical finding. REVISE on any high finding or any gate under 70. PUBLISH only above 80 overall.

Tell gate
weight 20 · script plus a manual pass over 62 tells
CRO gate
weight 30 · value proposition, one CTA, proof, friction, claims
GEO gate
weight 20 · crawlers, HTML content, schema, answer blocks, llms.txt
Performance & accessibility
weight 15 · budgets, Core Web Vitals, WCAG 2.2 checks
Trust gate
weight 15 · real proof, legal lines, honest pricing and guarantees
example output

Three pages. Three worlds. One method.

Each page below was built with PageSmith Studio from a one-page brief. Fictional businesses, real process: a palette from the physical world of the business, a typeface with a reason, one call to action, answer-first FAQs, structured data. Navigate them here or open them full screen.

fornobianchi.example Open full page

Three decisions only this bakery could get

  • The palette comes from the crust, the crumb and the oven stone, not from a template.
  • Fraunces, with its ink traps, because a wood-fired loaf is not a sans-serif.
  • The hero is the loaf on the stone, and the headline is the 8 a.m. promise.

All three pages pass the tell check at 100/100 and carry Organization, FAQPage and offer-type schema. Scores from scripts/tell_check.py, run on 20 October 2026.

Look first.

Design second.

Prove before you claim.

Ship only on PUBLISH.

Twelve modes, one command.

Type /page and a mode. Each mode loads only the playbook it needs, so your context stays small.

  1. /page brief

    Intake from a URL or eight questions.

    page/brief.md
  2. /page build

    The whole pipeline, end to end.

    page code + PAGE-NOTES
  3. /page design

    Research and three scored directions.

    design-directions.md
  4. /page copy

    Conversion copy for every section.

    copy.md
  5. /page geo

    SEO + GEO layer for an existing page.

    schema · llms.txt · robots
  6. /page gate

    Five gates, one verdict.

    gate-report.md
  7. /page audit

    Score any live URL, 0 to 100, six dimensions.

    audit.md + audit.json
  8. /page report

    White-label client report, print-ready.

    report.html
  9. /page proposal

    Three-tier offer written from the audit.

    proposal.md
  10. /page compare

    Re-audit a month later; show the points gained.

    compare.md
  11. /page fix

    Apply the plan to the codebase, one commit per finding.

    commits
  12. /page deploy

    Cloudflare, Vercel, Netlify, GitHub Pages, plus checks after.

    live URL

What's inside

Forty-seven text files. No binaries, no package installs, no environment variables. The only network calls go to the URL you name and to the deploy provider you choose.

What each script reads, fetches and writes
ScriptReadsNetworkWrites
page_audit.pynothing localonly the URL you pass, its robots.txt, llms.txt, sitemapaudit.json
tell_check.pythe path you passnonetells.json
schema_gen.pyyour briefnoneschema.jsonld
llmstxt_gen.pyyour briefnonellms.txt
report_build.pyaudit, confignonereport.html
 
pagesmith-studio/
├── SKILL.md            router: triggers, modes, principles, pipeline
├── playbooks/          12 playbooks, one per mode
├── agents/             researcher · designer · copywriter · geo-analyst · gatekeeper
├── scripts/            5 Python scripts, standard library only
├── references/         tells · copy rulebook · section library · CRO · GEO
│                       performance & a11y · typography & colour · scoring
├── templates/          brief.json · llms.txt · proposal · 6 JSON-LD types
├── examples/           a brief · an audit · a build session
├── white-label/        config.json for reports and proposals
├── README.md · LICENSE · CHANGELOG.md

Built for people who sell pages.

Audit in the first meeting. Report before you leave. Proposal the same day. Progress every month.

  1. /page audit scores the client's live page 0 to 100 across six dimensions, with evidence lines, a prioritised plan and a 30-day projection.
  2. /page report turns it into a print-ready A4 document with your logo, colours and footer. One JSON file to white-label everything.
  3. /page proposal writes the three-tier offer from the audit, priced with your rates.
  4. /page compare re-audits a month later and shows the client exactly how many points they gained.
Split illustration: a dark panel with the words VERDICT: PUBLISH and five gate bars, beside a bone-paper bakery page with a loaf on the oven stone
The gate verdict and the page it approved.

One price. Every project.

$25 one-time, per buyer

  • 12 modes, 12 playbooks, 5 subagent definitions
  • 5 scripts, Python standard library only
  • 62 tells with fixes, copy rulebook, section library, CRO checklist, GEO reference, performance budgets, typography and colour method
  • White-label report and proposal templates
  • Buyer licence: unlimited projects, your clients' projects included
  • Updates included; money-back guarantee handled by Agensi
Get it on Agensi

Runs inside your agent. Nothing to download, nothing to keep up to date.

Works with

Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, Windsurf, OpenClaw, Cline, Zed, and any other agent that reads SKILL.md.

Does not need

API keys, package installs, a build step, or a design degree. Python is optional. Chrome adds Lighthouse and PDF export.

Questions before buying

Is PageSmith Studio just another frontend-design skill?

No. Design-only skills stop at the look. PageSmith Studio runs research, three scored design directions, conversion copy under a claims policy, a static-first build, the SEO and GEO layer, and five quality gates with a PUBLISH, REVISE or HOLD verdict. It also audits live URLs, writes white-label client reports, three-tier proposals and before-and-after comparisons.

Which AI agents does PageSmith Studio work with?

PageSmith Studio works with any agent that reads SKILL.md: Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot, Windsurf, OpenClaw, Cline, Zed and others. Its five subagent definitions are used where the host supports subagents and ignored otherwise.

Do I need to install anything to use PageSmith Studio?

No. The core of PageSmith Studio works with your agent's fetch and editor. Python 3, standard library only, unlocks the five scripts; Chrome or Node unlock Lighthouse and PDF export. The skill never installs packages and never reads environment variables.

What does PageSmith Studio do with my data?

Nothing leaves your machine except requests to the URL you ask PageSmith Studio to audit, plus that site's robots.txt, llms.txt and sitemap, and requests to the deploy provider you choose. Scripts read and write only inside your project directory.

Can I use PageSmith Studio for client work and put my brand on the reports?

Yes. PageSmith Studio reports and proposals are white-label: one JSON file holds your name, logo, colours, footer and rates. The buyer licence covers unlimited projects, including your clients' projects, and everything the skill produces belongs to you.

Does PageSmith Studio guarantee rankings or AI citations?

No tool can. PageSmith Studio removes the known blockers, such as crawler 403s, missing structured data and no answer blocks, builds the content AI engines cite, and gives you a monthly measurement routine. Its scores are evidence-based estimates, not promises of traffic or revenue.

Stop shipping the page everyone has.

Install in thirty seconds. Research, three directions, copy, build, GEO layer, five gates. PUBLISH or it does not ship.

Get it on Agensi