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 flashcardsgit 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/flashcards)<a href="https://agentmods.dev/skills/androoagi/starnet/flashcards"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/flashcards/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/flashcards"><img src="https://agentmods.dev/badge/skills/androoagi/starnet/flashcards.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 | $0.00041 | $0.01292 |
| Opus 5.5 | $0.00016 | $0.00517 |
| Sonnet 5.5 | $0.00008 | $0.00258 |
| Haiku 4.5 | $0.00004 | $0.00129 |
Grade A, and why
flashcards 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.
Spaced repetition works only when two things are true: the cards ask for real recall, and the grading is honest. A deck graded kindly schedules itself too far out and the Commander forgets the material while the numbers say otherwise.
Method
- Get the material and the goal. Read the Commander's notes, chapter or document with
fs.read(or take pasted text). Ask what it is for — an exam date, a language, a job — and how many minutes a day they will study. That sets the deck size. - Write the cards. One fact per card. The question must force recall: no yes/no, no answer hidden in the wording, no "which of these". The answer is short — a word, a number, one sentence. A list of five things becomes five cards, or one card per missing item. Add the reverse card only when both directions matter (term to meaning, meaning to term). Every card records where in the source it came from.
- Let the Commander prune. Show the draft deck with a count. Cut cards they already know cold, reword the unclear ones, and flag anything in the source that looked wrong or thin instead of quietly fixing it.
- Save the deck as a file.
flashcards/<deck-name>.csvwith one row per card:id, question, answer, source, due, interval_days, ease, reviews, misses. New cards start withdue= today,interval_days= 0,ease= 2.5. Keep rows sorted bydue, earliest first, so the cards to study are always the top of the file. Quote fields that contain commas. Read the file back after writing it. - Run the session. Take today's date from the run; never guess it. Pick cards with
dueon or before today, oldest first, then at most ten new cards unless the Commander set another number. Ask ONE question, then stop and wait for the answer. No hints, and no answer shown before they reply. - Grade against the card, not against your own knowledge.
right= every fact in the card's answer is present and nothing in the reply contradicts it; spelling slips and different wording are fine.partly= some of it.wrong= missing, contradicting or "I don't know". When in doubt between two grades, give the lower one. Then show the card's answer and what was missed, in one or two lines. - Reschedule by rule. Wrong: interval 1 day, ease down 0.2 (never below 1.3), misses up one. Partly: interval 1 day, ease unchanged. Right: an interval of 0 becomes 1 day, 1 becomes 3 days, and anything longer becomes the last interval times ease, rounded; ease up 0.05 (never above 2.8). Every review adds one to
reviews, anddue= today + interval. Do the date sums incode.runrather than in your head. - Save as you go. Write the updated rows with
fs.writeevery few cards and at the end, so a dropped session loses nothing. If the write is refused because the file changed, read it again and re-apply. - Show progress. After the session: cards reviewed, right · partly · wrong, how many come due tomorrow and over the next seven days, and the "sticky" cards missed four times or more. Offer to split or reword sticky cards; change them only with the Commander's yes.
- Offer a daily reminder (optional). With
routine.create, set a routine at the Commander's study time whose prompt is complete on its own: read the deck files inflashcards/, count rows withdueon or before today, and reply with one line giving the count. Create it from the chat where the reminder should arrive and passdeliver: origin, so each reply lands there; a scheduled run cannot send withchannel.send, because a send needs an approval nobody is present to give. Tell it to reply exactly[SILENT]when nothing is due. A scheduled run cannot wait for answers, so it reminds; it never quizzes or regrades. Do not count oncode.runthere — that is why the deck stays sorted by date.
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 · 41 tokens per session scan A 8fa86d2c64d6
flashcards is a skill published in the GitHub repository androoAGI/starnet (1,468 stars, last pushed 3d ago), licensed MIT. It adds 41 tokens to every session and 1,292 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
hps-academic-paper
A review deck on compositional generalization in large language models — the field map, the gap, the evidence, and open questions. Built as a decision-grade academic research deck for PI, lab group, reviewers.
fs-notebook-tabs
A computer-science capstone: an on-device ML keyboard that predicts next words privately — problem, method, evaluation, and defense answers. Built as a decision-grade coursework defense deck for professor, defense committee.
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.
html-ppt-zhangzara-pin-and-paper
A field-biology capstone on urban pollinator decline — the survey design, the data, the contribution, and the caveats. Built as a decision-grade coursework defense deck for faculty reviewers.
html-ppt-zhangzara-cartesian
An economics senior thesis on the employment effects of local minimum-wage increases — identification strategy, evidence, and limitations. Built as a decision-grade coursework defense deck for thesis committee.