synthetic-sciences/openscience is an AI workbench that carries out scientific research by reading papers, forming hypotheses, writing and running code, conducting experiments, analyzing results, and preparing reports. Researchers use it for work in machine learning, biology, physics, and chemistry with remote or local models. Catalogue add-ons extend its scientific workflows through skills and instructions.
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 synthetic-sciences/openscience --skill computegit clone --depth 1 https://github.com/synthetic-sciences/openscienceWrote 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/synthetic-sciences/openscience/compute)<a href="https://agentmods.dev/skills/synthetic-sciences/openscience/compute"><img src="https://agentmods.dev/badge/skills/synthetic-sciences/openscience/compute/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/synthetic-sciences/openscience/compute"><img src="https://agentmods.dev/badge/skills/synthetic-sciences/openscience/compute.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 | $0.00127 | $0.01454 |
| Opus 5.5 | $0.00051 | $0.00582 |
| Sonnet 5.5 | $0.00025 | $0.00291 |
| Haiku 4.5 | $0.00013 | $0.00145 |
Grade A, and why
compute 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 19d 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 — 104 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Compute
The question is never "which cloud is best"; it is "which of the targets this user has configured fits this run, and what will it cost". Discover before you plan, size before you dispatch, and hand anything longer than a coffee to the job system so the session, the logs and the metrics outlive the shell.
Rules
- Targets first.
compute_job {"action":"targets"}lists what exists: local runtime, saved SSH and scheduler hosts, Modal when connected. Plan only against a listed target. A provider skill from the library (modal-serverless-gpu,runpod-gpu-cloud,lambda-gpu-cloud,vast-ai-gpu-cloud,skypilot-multi-cloud-orchestration,together-ai-inference,fireworks-ai-inference,tinker-fine-tuning; the core index lists them) is loaded only when that provider is configured or the user asks to set it up; never plan work for compute the user does not have. - Local shell for seconds, kernel for interactive state, job for everything else. A
command that may exceed a few minutes, needs a GPU, or must survive a disconnect goes
through
compute_job start; a job that settles within the dispatch's grace returns its outcome and output tail in that step, a longer one wakes you when it ends. Never a shellsleeploop, and nowaitwhile other work can proceed. - Size and price before dispatch. Estimate wall-clock from a short local or scaled-down
run, choose the smallest instance that fits memory and time, and state the estimate
(provider, GPU, duration, price) in the approval. Remote jobs are approved against an
immutable plan digest;
compute_job planpreviews it. - Declare outputs. Artifacts and checkpoints are relative paths under the job
workspace, listed in the request, so they are delivered and survive. Nothing important
is written to
/tmpor printed only to the terminal. - Track metrics. Training and evaluation scripts import
openscience_track(orwandb, shimmed inside a study) so curves and final numbers land in the Autoresearch pane and the experiments tool instead of a log to grep. - Checkpoint long runs and make them resumable from the declared checkpoint path; the job system can be interrupted by the user, a budget, or a kill rule.
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.
- 19d ago Changed · +2 lines 20ec98f472a9
- 1mo ago Changed · +6 lines 9a809780ee6f
- 1mo ago First seen · 96 lines · 127 tokens per session scan A e2e00cff0346
compute is a skill published in the GitHub repository synthetic-sciences/openscience (4,004 stars, last pushed 3d ago), licensed Apache-2.0. It adds 127 tokens to every session and 1,454 once invoked, about $0.0005 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 2026-09-14.
Other skills, from other repositories
serving-llms-vllm
Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.
architecture-diagram
Dark-themed SVG architecture/cloud/infra diagrams as HTML.
songsee
Audio spectrograms/features (mel, chroma, MFCC) via CLI.
arxiv
Search arXiv papers by keyword, author, category, or ID.
research-paper-writing
Write ML papers for NeurIPS/ICML/ICLR: design→submit.
paper-revision-author
Revise independently drafted paper sections into one coherent LaTeX body before the abstract is written.