StarNet is a local-first desktop harness for creating AI agents and arranging them in a pixel-art space station whose layout represents their capabilities, permissions, and workflow. It lets people run real model calls and tools through multiple bounded agents, using their own provider keys or accounts and optionally messaging the station through chat services. The catalogue entries are skills that extend or support work performed by StarNet agents.
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 skills add androoAGI/starnet --skill conversion-reviewgit clone --depth 1 https://github.com/androoAGI/starnetWrote 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/androoagi/starnet/conversion-review)<a href="https://agentmods.dev/skills/androoagi/starnet/conversion-review"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/conversion-review/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/androoagi/starnet/conversion-review"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/conversion-review.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.00047 | $0.01969 |
| Opus 5.5 | $0.00019 | $0.00788 |
| Sonnet 5.5 | $0.00009 | $0.00394 |
| Haiku 4.5 | $0.00005 | $0.00197 |
Grade A, and why
conversion-review 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 12d 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 — 126 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Conversion Review
Review a marketing page or a form and say why visitors are not taking the action, and what to change first. The bar: you looked at the real page at desktop and phone widths, every finding points to something visible on the page, and you never promise a lift you cannot back.
When to use
- "Why isn't this page converting?" "Review my landing page." "People give up on this form."
- A URL with "any feedback?"
- Homepage, landing page, pricing page, feature page, blog post, lead or contact form.
Not for writing the page from scratch (landing-copy), testing ad messages (ad-copy-testing), being found in search (seo-audit), or deciding the prices themselves (pricing-strategy).
Method
- Frame the review. Establish the page type, the one action the page exists to get (sign up, book a demo, buy, subscribe, download, contact), and where visitors come from (ad, email, search, social), because the page must match the promise that sent them. Ask for the current conversion rate and any visitor research the Commander has (heatmaps, recordings, survey answers); carry on without them. Look for a product brief in the workspace with
fs.search. - See the page at desktop width.
browser.navigateto the URL. Take abrowser.screenshotbefore scrolling: that first screen is what every visitor sees. Open the saved image withfs.read. Pull the full copy withbrowser.get_textand the list of buttons, links and form fields withbrowser.snapshot. If you need the whole page as one document,browser.pdfsaves it. - See it at phone width. Resize with
browser.viewport(width 375, height 812,mobile: true). It is a deferred tool: load it withtool.searchif it is not in your list. Take a freshbrowser.snapshotandbrowser.screenshot. Check that the headline and main button show without scrolling, the text is readable, nothing needs sideways scrolling, buttons are large enough to tap, and the form is usable. Set the width back (1440 by 900) when done. If the width cannot be changed, say the phone check was not done. - Run the five-second test. From the first-screen screenshot alone, answer: what is this, who is it for, what should I do next? Write your answers before reading the rest. If you cannot answer, that is the top finding.
- Work through the seven dimensions below, in order. They are ranked by impact.
- Review the form, if the page has one, against
references/forms.md(included at the end of this skill). Count the fields. Do not submit it. - Check for breakage.
browser.networkshows failed and very heavy requests. A broken image or a slow first screen is a conversion problem too. - Write the review in the output format, save it with
fs.write, and name it withdeliverable_note.
What ships with it
4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 12d ago First seen · 126 lines · 47 tokens per session scan A 38d542c604f7
conversion-review is a skill published in the GitHub repository androoAGI/starnet (1,468 stars, last pushed 3d ago), licensed MIT. It adds 47 tokens to every session and 1,969 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-10-04.
Other skills, from other repositories
fs-electric-studio
OpenDesign as an enterprise design platform: a buyer-forwardable proposal for a design-org's economic buyer — pain, value, ROI, rollout. Built as a decision-grade B2B sales deck for economic buyer, design VP, procurement.
hps-true-blueprint
OpenDesign's engineering blueprint: how the sandbox, sidecar, and daemon fit — the system diagram and the invariants. Built as a decision-grade product management deck for engineering org.
ve-terminal-mono
OpenDesign from the CLI: driving the full design workflow with the od command — scripted, composable, agent-ready. Built as a decision-grade AI literacy deck for developers, power users.
kami-landing
Produce a print-grade single-page kami (紙 / 纸) document — warm parchment canvas, ink-blue accent, serif at one weight, no italic, no cool grays. The output reads like a professional white paper or studio one-pager, not an app UI. Multilingual by design (EN · zh-CN · ja). One self-contained HTML file, zero dependencies.
hps-bauhaus
A visual-design fundamentals course for new brand designers — grid, type, color, and the exercises that build the eye. Built as a decision-grade professional training deck for junior designers, new hires.
hps-memphis-pop
A pop-culture retrospective on how 1980s design language shaped today's apps — the scenes, the turning point, and the takeaway. Built as a decision-grade story deck for talk audience, design community.