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 literature-reviewgit 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/literature-review)<a href="https://agentmods.dev/skills/synthetic-sciences/openscience/literature-review"><img src="https://agentmods.dev/badge/skills/synthetic-sciences/openscience/literature-review/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/literature-review"><img src="https://agentmods.dev/badge/skills/synthetic-sciences/openscience/literature-review.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 | $0.00123 | $0.01763 |
| Opus 5.5 | $0.00049 | $0.00705 |
| Sonnet 5.5 | $0.00025 | $0.00353 |
| Haiku 4.5 | $0.00012 | $0.00176 |
Grade A, and why
literature-review 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 1mo 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 — 120 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Literature review
You are the retrieval ranker. Run the searches, look at every candidate, decide what is worth reading, and read it before it becomes a claim. Never hand the loop to a worker: the ranking is the judgement, and a worker without the conversation ranks for a different question.
Non-negotiables
- One window, one priority, fixed before the first query. Recency, historical or default; a date bound only if the question has one. Never widen the window mid-loop to find "the famous paper" the user excluded.
- Budget by difficulty. Rate the question 1–10. Difficulty 1–3 gets the initial round only; 4–7 gets one follow-up round; 8–10 gets two. A round targets one concrete gap (an acronym, a method, a benchmark, a venue), never a rephrasing.
- Deduplicate by identifier, then by DOI or arXiv id, then by normalized title. Prefer the published version's record; note when the arXiv version is the one read.
- Rank by topical fit. The APIs already blend recency and citations into their order; do not apply those preferences a second time, and never compare citation counts across sources.
- Read before you claim. A candidate list may link titles; a finding attributed to a paper requires having read at least its abstract, and its results section for a number. Put the source link right after the sentence it supports.
- Stop when coverage is sufficient. Five to fifteen strong candidates beat forty padded ones. The budget is a cap, not a target.
- Size the review to the request. "A bit of a review", "a quick look", "what is the closest work" is discovery mode: two or three searches, three to five papers read, the concrete missing comparison named, one follow-up search only if that gap matters. The citation ledger, corpus downloads and exhaustive screening are for requests that ask for them. Report substantive findings and blockers, not each retrieval step.
Workflow
- Frame: the question in the user's terms, the window, the priority, the difficulty.
- Initial round: two or three targeted
literature searchcalls, concurrently. - Deduplicate, rank, and decide whether a follow-up round is warranted.
-
literature readthe 3–5 load-bearing papers (more only if asked). - Write the synthesis with claim-level links; list what was excluded and why.
What ships with it
2 files 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.
- 1mo ago Changed · +5 lines 5c76d18c1e70
- 1mo ago First seen · 115 lines · 123 tokens per session scan A 242d1b930508
literature-review is a skill published in the GitHub repository synthetic-sciences/openscience (4,025 stars, last pushed today), licensed Apache-2.0. It adds 123 tokens to every session and 1,763 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
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.
paper-plot-stub
Plot a results CSV (x, ybaseline, yours) as a two-line matplotlib chart and write a PDF. Demo-only.
knowledge-extension-ecosystem-lsp-code-intelligence
A codebase guide to LSP, the standard protocol behind editor features such as symbol queries, file diagnostics, and language-aware code information.