Systems Lab

Agent skill

microsite-builder

Generates the personalized microsite for one account from a research packet.

activeSelf-containedInstructions only688 words

Filed under Content and SEO.

From V00nikos/microsite-factory · 4 skill entries · 1 · pushed 2026-09-11

What it does when it runs

Generates the personalized microsite for one account from a research packet. Use whenever a pipeline task requires building, generating, or regenerating a microsite, GTM page, landing page, or outreach page for a target account. Takes the account-researcher JSON packet plus the operator's positioning memory; outputs a single self-contained HTML file with the tracking beacon embedded. Contains the mandatory storytelling spine and design language — always load this skill before writing any microsite HTML, never freestyle a page.

Automated analysis of the skill and the 0 files bundled beside it. A skill’s own description is written to be selected by an agent, so it describes the job and not the dependencies.

Keys and connectors you must supply
None found.
Hosts it reaches
No third-party host appears in the skill or its bundled files.
Tool permissions it declares
No allowed-tools in the frontmatter. It only issues instructions, so there is nothing to bound.
Actions present in the files
None. Instructions only.

Ask about microsite-builder

Opens your assistant with this page's verified links already in the prompt.

Is this safe to install?ClaudeChatGPT
Adapt it to my stackClaudeChatGPT
What else do I need for it to workClaudeChatGPT
Rather ask a human? Talk to Cheetah
git clone --depth 1 --filter=blob:none --sparse https://github.com/V00nikos/microsite-factory.git /tmp/microsite-factory
git -C /tmp/microsite-factory sparse-checkout set "skills/microsite-builder"
mkdir -p ~/.claude/skills/microsite-builder
cp -R "/tmp/microsite-factory/skills/microsite-builder/." ~/.claude/skills/microsite-builder/

Picked up without a restart. A project skill of the same name is shadowed by your personal one. For one repository only, swap ~/.claude/skills for .claude/skills. Claude Code docs ↗

The folder is the same in every client that implements the format — 46 of them — so if yours is not above, only the destination changes.

Reproduced in full from V00nikos/microsite-factory/blob/05c2a30773cf42f8621435dc35b9deb9882e727c/skills/microsite-builder/SKILL.md, which is licensed MIT (repository). 688 words, 6 headings.

Microsite builder

You build one page per invocation. Inputs: the research packet (JSON), the operator's positioning (from memory: one-liner, proof points, tone rules, forbidden claims), and the design system assignment. Output: ONE self-contained HTML file, no external framework, loads in under 1 second.

Step 1 — choose the angle (storytelling)

Pick exactly one archetype based on the strongest findings (hook_strength) and the likely_objection:

ArchetypeUse when findings show…
costfunding pressure, efficiency hires, consolidation news
speedaggressive roadmap, competitive launches, eng hiring spikes
riskregulatory vertical, security posture, compliance signals
talenthard-to-fill roles open long, team scaling pains
competitiona named rival moved recently

Log the choice: {"angle": "...", "because_findings": [indices], "objection_addressed": "..."}. This log is what the operator reviews and regenerates against — the angle IS the editable surface of this product.

Step 2 — the narrative spine (fixed order, five sections max)

  1. Their world — hero states a thesis about THEIR business in the first line, built from the top finding. Their company name in text. Not about us.
  2. Their pressure — 2–3 sourced findings woven into one tension. Every claim carries a visible citation link. Citations are the anti-spam proof.
  3. The bridge — the operator's product enters ONLY here, framed as the specific answer to the pressure above. Use positioning memory verbatim for claims; never invent capabilities.
  4. One proof — a single proof point from positioning memory, chosen to match the reader lens (technical proof for CTO, ROI proof for CFO).
  5. One CTA — named for what happens: "See the 20-minute teardown for {Company}", never "Learn more". One CTA only.

Copy rules: active voice, sentence case, no filler adjectives, the page speaks about them for the first two-thirds. Echo their_phrase from brand hints once, naturally. Word budget: 250–400 words total. Dense pages read as brochures.

Step 3 — design language (mandatory)

Derive, don't decorate. The page's identity comes from the TARGET's world via brand_hints, mapped to one of four fixed design systems:

SystemRegisterDisplay faceBody faceCharacter
ledgerprecise/serious (finserv, legal)FrauncesInterdisciplined grid, generous white, one hairline accent in their hue
terminaltechnical (devtools, infra)JetBrains MonoSöhne-like grotesk (fallback: Inter)mono display, code-block signature, dark-on-light NOT dark theme
cliniccalm/clinical (health, edtech)Source Serif 4Source Sans 3soft radius, muted palette, spacious line-height
signalbold (consumer-ish B2B, martech)Space GroteskInteroversized display type, one saturated accent from their hue

Load faces from Google Fonts with display=swap; subset to used weights (400/500 only).

One signature element per page — spend all boldness in exactly one place, derived from the research: an inline diagram of THEIR stack, a small interactive before/after of THEIR workflow, a live-feeling counter tied to THEIR metric. Everything else stays quiet. No signature = QA fail.

Banned looks (auto-fail): warm-cream bg + high-contrast serif + terracotta accent; near-black bg + acid green; hairline broadsheet with zero radius and dense columns. These are AI tells. If your draft resembles one, redesign.

Floor (non-negotiable): responsive to 360px, visible keyboard focus, prefers-reduced-motion respected, semantic headings, single HTML file, no logo or trademark styling of the target — their name in text only, with a "Prepared for {Company} by {Operator}" line in the footer and a contact link for takedown requests.

Step 4 — impressions beacon (hooks tracking)

Embed exactly this instrumentation before </body>, pointing at the fleet state endpoint (Convex HTTP endpoint or the API server URL passed in config):

<script>
(function(){var e="{BEACON_URL}",a="{ACCOUNT_ID}";
function s(t,d){try{navigator.sendBeacon(e,JSON.stringify({a:a,t:t,d:d||null,ts:Date.now()}))}catch(_){}}
s("view");var f=!1;addEventListener("scroll",function(){if(!f&&scrollY>document.body.scrollHeight*0.5){f=!0;s("scroll50")}},{passive:!0});
document.addEventListener("click",function(ev){var c=ev.target.closest("[data-cta]");if(c)s("cta")});
})();
</script>

Tag the CTA anchor with data-cta. Three events only — view, scroll50, cta. Do not add fingerprinting, identity capture, or extra trackers; visits are intent signal, not surveillance.

Output contract

{"html_path": "sites/{account_id}/index.html",
 "angle_log": {...},
 "design_system": "ledger|terminal|clinic|signal",
 "signature_element": "one-line description",
 "word_count": 0,
 "claims_cited": 0}

Other skills for the same job

Different authors, same problem. Matched on the words in the skill name, across every library in the catalogue except this one.

Need help setting it up?

This page tells you what microsite-builder does and what it needs. Cheetah builds the agent setup it runs inside: data, CRM, sequencing and the guardrails.

Book a call →

The directory stays free. There is nothing gated behind this.