Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/onewave-ai/open-agent-stack/site-buildernpx skills add OneWave-AI/open-agent-stack --skill site-buildergit clone --depth 1 https://github.com/OneWave-AI/open-agent-stackWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/onewave-ai/open-agent-stack/site-builder)<a href="https://agentmods.dev/skills/onewave-ai/open-agent-stack/site-builder"><img src="https://agentmods.dev/badge/skills/onewave-ai/open-agent-stack/site-builder.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00058 | $0.01161 |
| Opus 5 | $0.00029 | $0.00580 |
| Sonnet 5 | $0.00012 | $0.00232 |
| Haiku 4.5 | $0.00006 | $0.00116 |
Grade A, and why
site-builder scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 33 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Site Builder
A marketing site is not a small app. Apps are judged on what they do; sites are judged in about four seconds on what they say. Which means the failure modes are different: an app breaks, a site just quietly fails to convince anyone and nobody tells you why.
Two rules govern everything below. Copy comes before layout -- designing a hero before knowing the sentence is how you end up with a headline written to fit a box. And the site does one job -- the visitor should never wonder what they are meant to do next.
Workflow
- Answer the four questions. Before a single file: what does this sell, to whom, what does the visitor do next, and why should they believe it. If the user cannot answer the fourth with something concrete -- a name, a number, a screenshot, a quote -- that is the site's real problem, and it will not be solved by design. Say so.
- Inventory the pages. Most sites need fewer than the user thinks: home, one page per real offering, about, contact, plus legal. A five-page site done properly beats a twelve-page site with eight thin pages -- thin pages hurt both conversion and search. Decide the nav now, because nav length constrains everything after.
- Write the copy first, as plain text. Per page: the one-line promise, the proof, the objection it answers, the call to action. Then read it aloud. The headline states what the thing is and who it is for -- clever comes second, and only after clear. Specific beats sweeping every time: "invoices chase themselves" over "streamline your workflow". No "Unlock", no "Supercharge", no "In today's fast-paced world".
- One conversion path. Pick the single action that matters -- book, buy, sign up, reply -- and make everything point at it. The same CTA wording throughout, present above the fold and repeated at natural decision points. Every competing secondary action you remove raises the odds of the primary one. If there is a form, it asks the fewest fields you can actually act on.
- Build the page in sections, not as a wall. Vary the rhythm: full-bleed, then narrow, then asymmetric. Let one element be genuinely large. Standard order works and there is no prize for reinventing it -- hero, proof, what it does, how it works, objection handling, CTA -- but every section earns its place by doing a job in that sequence. See
polish-passfor stripping the template smell out of the result. - Do the metadata for real. Per page: a unique title under ~60 characters, a description that would make someone click, canonical URL, and an OG image that renders correctly -- test the actual link preview, do not assume. Site-wide:
sitemap.xml,robots.txt, favicon at every size, and JSON-LD (Organization, plusLocalBusinessif there is an address). A site without OG images looks broken every time anyone shares it, which is exactly when it matters most. - Make the forms deliver. The most common failure in this entire skill: the contact form that resolves into nothing. It must reach a human's inbox or a database, confirm submission visibly, handle its own error state, and have spam protection that is not a hidden field a bot fills in anyway. Then submit it yourself, on production, and confirm the message arrived. See
make-it-real. - Budget the performance. Marketing sites are mostly images, and images are the whole performance story. Modern formats, correct dimensions, lazy below the fold and priority above it, and explicit width and height so nothing shifts. Fonts subset and preloaded,
font-display: swap, two families maximum. Target LCP under 2.5s on a throttled mobile connection -- measure it, because desktop tells you nothing. - Verify on a phone first. Most traffic is mobile and most of these sites are built in a desktop window. Check the hero at 390px wide, tap targets at 44px, that nothing scrolls sideways, and that the nav works with a thumb.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 33 lines · 58 tokens per session scan A 22c50447b0af
site-builder is a skill published in the GitHub repository OneWave-AI/open-agent-stack (2 stars, last pushed 23d ago), licensed MIT. It adds 58 tokens to every session and 1,161 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
dcf-model
Build discounted cash flow valuation workbooks in Excel.
baoyu-article-illustrator
Article illustrations: type × style × palette consistency.
openhands
Delegate coding to OpenHands CLI (model-agnostic, LiteLLM).
agentmail
Use when an agent needs AgentMail CLI email inboxes.
mcporter
List, auth, and call MCP servers/tools from the terminal.
mem0-vercel-ai-sdk
Mem0 provider for Vercel AI SDK (@mem0/vercel-ai-provider). TRIGGER when: user mentions "vercel ai sdk", "@mem0/vercel-ai-provider", "createMem0", "retrieveMemories", "addMemories", "getMemories", "searchMemories", "mem0 vercel", "AI SDK provider", "AI SDK memory", or is using generateText/streamText with mem0. Also…