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 spend-aware-workgit 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/spend-aware-work)<a href="https://agentmods.dev/skills/androoagi/starnet/spend-aware-work"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/spend-aware-work/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/spend-aware-work"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/spend-aware-work.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 | $0.00050 | $0.00928 |
| Opus 5.5 | $0.00020 | $0.00371 |
| Sonnet 5.5 | $0.00010 | $0.00186 |
| Haiku 4.5 | $0.00005 | $0.00093 |
Grade A, and why
spend-aware-work 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 — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Metered runs spend real money on the Commander's account. The station has rails that stop a run when a cap is hit — but a run stopped by a cap has usually already wasted what it spent. Plan so you never meet the rail.
The rails you work inside
- The day rail. A soft cap on metered spend per day ($25 by default, editable in SETTINGS > SPENDING LIMITS as PER DAY, 0 = off). Hitting it ends the run with reason
budget, and the Commander can RESUME with one click. Per-run, per-agent and whole-station caps can be set there too. Subscription sign-ins have no dollar meter and never touch it. - Line budgets. Every line has a LINE BUDGET: stages after the first (6 by default), dollars per message ($2 by default) and an optional dollars per day. Real test runs count against it; WATCH IT is free.
- Crew. Every dispatched worker runs its own loop and spends on its own model.
Method
- Read the evidence you have. station.layout shows each line's budget and its spend today; routine.list shows which routines run, how often and on which model. Each team.dispatch worker result reports what it cost. Past runs are the best estimate of the next one.
- Estimate before you start. Count the expensive units: model turns, workers, pages read, images and voice takes made. Multiply by what similar runs cost. When the job is bigger than an ordinary chat, state the estimate in dollars and say which part dominates it.
- Right-size the model. An agent runs on the model pinned in its Dossier CONFIG, or the station default. For standing work, pin a cheaper model per routine or loop (
model/provideron routine.create, routine.manage and loop.create) when the job is extraction, formatting or triage; keep the strong model for judgment. For images, stay on the default model and use the premium one only where readable text must appear. - Right-size the tools. Read a page with web_fetch before driving a browser. For text-only scraping, browser.intercept (find it with tool.search) blocks images, media and fonts. Use code.run to loop over read tools and return one compact answer instead of spending many turns.
- Fan out only on real seams. Five parallel workers cost about five times one. Dispatch when the pieces are truly independent; otherwise do the job in one run.
- Write stop rules into the plan. "Stop after three sources agree." "At most two retries per failing step, then report." Put the same limits into routine prompts and worker briefs, because unattended runs cannot ask.
- Stop and ask before crossing. When the work will clearly overrun the estimate, stop at a clean point, report what is done and what the rest will cost, and let the Commander choose.
What ships with it
1 file 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 · 40 lines · 50 tokens per session scan A 1a05b8715d00
spend-aware-work is a skill published in the GitHub repository androoAGI/starnet (1,468 stars, last pushed 3d ago), licensed MIT. It adds 50 tokens to every session and 928 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
orbit-general
Open Orbit briefing skill — selected by the Orbit pipeline when the user has two or more connectors connected. Pulls the past 24 hours of activity from every authenticated connector (GitHub, Linear, Notion, Slack, 飞书, Calendar, Gmail, Drive, Sentry, Vercel, …) and renders a single adaptive bento-grid dashboard at the…
orbit-gmail
Open Orbit briefing skill — selected by the Orbit pipeline when Gmail is the user's only connected connector, or when the user explicitly scopes their daily digest to Gmail. Pulls the past 24 hours of inbox activity (replies awaited, mentions, cc, auto- categorized bulk) from the user's authenticated Gmail connection…
orbit-notion
Open Orbit briefing skill — selected by the Orbit pipeline when Notion is the user's only connected connector, or when the user explicitly scopes their daily digest to Notion. Pulls the past 24 hours of document edits, comments, mentions, and database row changes from the user's authenticated Notion connection and…
discovery-question-form
Structured clarification form for unresolved material requirements.
cron
Use this skill only for scheduled or recurring tasks. Manage jobs with qwenpaw cron list/create/get/state/update/pause/resume/delete/run, and always pass --agent-id explicitly.
cron
A procedure for creating and managing tasks that run automatically at a future time or repeat on a schedule. It uses QwenPaw's scheduling commands and requires an agent identifier.