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 crew-handoffgit 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/crew-handoff)<a href="https://agentmods.dev/skills/androoagi/starnet/crew-handoff"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/crew-handoff/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/crew-handoff"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/crew-handoff.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 | $0.00878 |
| Opus 5.5 | $0.00019 | $0.00351 |
| Sonnet 5.5 | $0.00009 | $0.00176 |
| Haiku 4.5 | $0.00005 | $0.00088 |
Grade A, and why
crew-handoff 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 — 39 lines — stays where its author put it; the contents beside it link to each section on GitHub.
A delegated worker runs its own loop on your brief and nothing else. It cannot see your conversation or read your intent. Everything it needs to finish goes in the handoff — or it returns something plausible instead of something right.
Method
- Pick the crewmate by what they own (team.config). Read the candidate's Dossier — purpose and standing orders — and give the job to the specialist whose desk it is, not whoever is idle. If nobody fits, team.summon can add a specialist; the Commander confirms it in APPROVAL mode.
- Know what a worker can do. A dispatched worker runs unattended with web, files, memory and studio tools, but never the shell: keep shell work for yourself. station.layout shows who sits where.
- Write the prompt as the job, finished. One paragraph: the deliverable, who it is for, the scope boundary (what NOT to touch), and done stated as something checkable — "three options under $50, with links and the date checked", not "look into pricing".
- Put what you already know in
context. File paths, findings, decisions, constraints, the Commander's stated preferences. It arrives as settled starting knowledge, so the worker does not re-derive it. Do not restate the task there. - Pre-answer the questions they would ask. Walk the brief as the worker would: which source wins on conflict? what format? what if the data is missing? Answer each, or name the default to take and ask them to flag it.
- Fix the return shape. Say exactly what comes back. When the result feeds code or another step, pass a
resultSchema: the station validates the result, allows one bounded repair, and never accepts invalid output as done. - Choose where it runs. Pass
sessionwith an existing session's name to file the work there — session.list shows the real names, and an unknown name is refused, not guessed. Passbackground: truewhen you have other work meanwhile. - Dispatch (team.dispatch) with the worker's agentId, prompt, context and schema. In APPROVAL mode the Commander is asked first.
- Watch without hovering. For a background worker, team.subagents shows its status and event tail. Correct course with team.steer (id plus the current generation), stop it with team.interrupt, and restart a stale or failed one with team.resume after inspecting it.
- Check the return against your definition of done. The result's artifacts list is the proof of what the worker saved. Its files live in its own workspace — you cannot fs.read them, so never "verify" by looking in yours. Refer to them as "'s workspace: ".
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 · 39 lines · 47 tokens per session scan A deed41c0736b
crew-handoff 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 878 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
flowai-live-dashboard-template
Team-management dashboard skill in the FlowAI aesthetic — three tabs (Team Members, Team Details, Activity Log), KPI stat row, member table, role distribution bar chart, online presence and activity sparklines, and a top-contributors panel, all in a single self-contained HTML file with light/dark theming, hoverable…
weekly-update
OpenDesign's weekly metrics standup: this week's numbers, the one anomaly, and the single decision it forces. Built as a decision-grade data & finance deck for ops & growth team.
code-import
Read an existing repository's structure into the project cwd as a normalised snapshot the agent can analyse without re-walking the tree on every turn.
pm-spec
A product-specification page skill for documenting a feature's problem, goals, user stories, success measures, rollout plan, and open questions. A PRD is a document that explains what a product feature should achieve and why.
team-okrs
OKR tracker page — quarter banner, three objectives with their key results as progress bars, owner avatars, status pills, and a "this quarter at a glance" sidebar. Use when the brief mentions "OKRs", "key results", "objectives", or "目标".
huashu-keynote-black
OpenDesign's all-hands: the year in review, the three priorities, and what every team owns next quarter. Built as a decision-grade corporate strategy deck for whole company.