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 routine-craftgit 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/routine-craft)<a href="https://agentmods.dev/skills/androoagi/starnet/routine-craft"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/routine-craft/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/routine-craft"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/routine-craft.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 | $0.00000 | $0.00959 |
| Opus 5.5 | $0.00000 | $0.00384 |
| Sonnet 5.5 | $0.00000 | $0.00192 |
| Haiku 4.5 | $0.00000 | $0.00096 |
Grade A, and why
routine-craft 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 routine is a saved prompt the scheduler fires on its own. Nobody reads it at run time, nobody answers its questions, and it runs again tomorrow whether or not today's run was any good. Write it for that reader.
Method
- Check what exists (routine.list). It shows every routine, its schedule, last status and last error, and whether the scheduler is armed or stopped by the E-STOP. If a near-identical routine exists, improve it with routine.manage action=update instead of adding a twin — the station refuses duplicates and never recreates a routine the Commander deleted.
- Write the prompt as a complete brief. It is the only context the run gets. Include the goal, the sources or files to use, the bar for "worth reporting", the output shape, and where to save anything durable. No "as discussed", no "like last time".
- Give it a memory if it compares runs. Inside a scheduled run, routine.notepad (read, then write) keeps a private scratchpad for that routine alone — up to 8,000 characters, surviving restarts. Store the baseline (last value seen, last item id), never a log.
- Build in the silent path. Tell the routine: if nothing crossed the bar, reply exactly
[SILENT]. A silent run delivers nothing and pings no one; a failed run always reports. - Pick the schedule for the job, not the habit. The tool takes forms like
every 6h,0 9 * * 1-5,in 2h, or an ISO time. Passtimezonewith a cron expression. UserepeatTimesfor a routine that should end. - Choose the agent and model on purpose. Pass
agentId(oragentHint) so the routine runs on the right specialist. Pinmodel/providerwhen a cheaper model does the job. - Decide where results land.
deliver: localkeeps results on the station;deliver: originreturns each result to the chat or session the routine was created from. SetattachToSessionwhen replies should continue from the result. To feed one routine from another, usecontextFrom, withmonitorModeto run only when the upstream output changed. - Name what it cannot do alone. A routine that needs the terminal needs the Commander to grant that in the ROUTINES panel, and rewriting its prompt later clears the grant. Unattended runs are time-bounded (eight minutes by default), so size the job to fit.
- Create it (routine.create) and read the answer. Report schedulerArmed, schedulerHalted and the next run time exactly as returned.
- Prove one run. routine.manage action=run_now queues the routine for the scheduler's next tick; it does not run it now. Check routine.list afterwards for lastStatus and lastError, then read the result where
deliversays it lands.
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 · 0 tokens per session scan A 22897198f9cc
routine-craft is a skill published in the GitHub repository androoAGI/starnet (1,468 stars, last pushed 3d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 959 tokens. 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
session-librarian
Organize sessions by prompt: find, rename, archive, prune.
folyamatos-ellenorzes
Ellenorizd: 1) Naptar - van-e meeting 1 oran belul? 2) Email - jott-e surgos level az elmult oraban? 3) Kanban - van-e mai hataridovel kartya? Ha BARMIT talalsz ami fontos, szolj a beallitott csatornan tomoren. Ha minden csendes, ne irj semmit.
feishu-calendar
A Feishu calendar and scheduling skill. It manages calendars and events and can check when people are available or busy.
automation-architect
Workflow automation design using Zapier, Make, and n8n covering trigger-action architecture, multi-step automation workflows, error handling strategies, personal versus business automation patterns, ROI calculation, and a library of common automation recipes. Use when the user asks about automation architect, related…
vault-setup
Set up a machine to work with a workhub vault - check/install required software (git, Node.js, Claude Code, OpenCode), initialize the vault if needed, wire up Claude plugins, and run the OpenCode sync scripts. Use when onboarding a new machine, after a fresh vault-init, or when the harness feels half-configured.
wecomcli-todo
A tool for managing to-do items in WeCom, the workplace messaging and collaboration app used by many Chinese companies. It handles existing WeCom tasks and their titles, descriptions, people, status, and deadlines.