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 framework-topologygit 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/framework-topology)<a href="https://agentmods.dev/skills/synthetic-sciences/openscience/framework-topology"><img src="https://agentmods.dev/badge/skills/synthetic-sciences/openscience/framework-topology/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/framework-topology"><img src="https://agentmods.dev/badge/skills/synthetic-sciences/openscience/framework-topology.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 | $0.00128 | $0.01682 |
| Opus 5.5 | $0.00051 | $0.00673 |
| Sonnet 5.5 | $0.00026 | $0.00336 |
| Haiku 4.5 | $0.00013 | $0.00168 |
Grade A, and why
framework-topology 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 20d 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 — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Framework topology from a crystal structure
A framework's topology is a property of a graph you construct, and every wrong answer in this area is a wrong construction: a node that should have been two, an edge counted twice, a guest kept, a symmetry image missed, two independent nets read as one. Build the graph explicitly, print what it is at each stage, and identify the net from the graph's own invariants rather than from what the chemistry "usually" gives.
Procedure
- Expand the structure completely. Apply every symmetry operation to the asymmetric unit, merge atoms that coincide modulo the lattice (special positions), and keep the lattice vectors. Work in fractional coordinates with an explicit image vector per neighbour, so that a contact to a molecule in the next cell is a different edge from a contact to the same molecule in this cell.
- Find the molecules. Connect atoms by covalent bonding (covalent radii plus a tolerance, or the CIF's bond list) across periodic boundaries; the connected components are the molecules. Classify each: framework former, counter-ion, or guest (solvent, template, anything not part of the connected framework). Guests are removed before the framework graph is built; say which formulas were removed and why.
- Define the nodes the question defines. The default node is one framework molecule (its centroid); a task may instead name secondary building units, metal clusters, or the finite supramolecular clusters that hydrogen-bonded terminal groups form (a cluster of charged or acidic end groups shared by several molecules is a node in its own right when its members are not otherwise connected). Two-connected molecules are linkers, not nodes: contract them into edges before reading the net. State the node definition in the report; the RCSR symbol depends on it.
- Edges are distinct neighbours, with the question's contact criterion. Apply the criterion exactly as stated (donor and acceptor element sets, D···A distance, D–H···A angle; or coordination bonds) at the atom level, then collapse: two molecules joined by any number of qualifying contacts share one edge. A carboxylic-acid dimer, a bifurcated hydrogen bond or a chelating pair is one edge. Connectivity is the number of distinct neighbouring nodes counting periodic images separately, never the number of hydrogen bonds. A molecule with four dimer partners is 4-connected, not 8-connected.
- Build the periodic quotient graph (labelled quotient graph): nodes in one cell, edges labelled with the lattice translation between the two ends. Check its rank: the translations spanned by cycles must be three-dimensional for a 3-periodic net; rank 2 is a layer, rank 1 a chain, rank 0 a finite cluster (then the node definition or the contact criterion is wrong for a "framework").
- Interpenetration. Count the connected components of the periodic graph that are themselves 3-periodic: that number is the degree of interpenetration, and each component is analysed as its own net (they are usually identical). Components that are finite are guests or clusters that should already have been handled in steps 2–3. A single component means "not interpenetrated" (degree 1); report the degree as the question defines it.
- Identify the net. Compute the coordination sequence of every node class (the
number of nodes at graph distance 1, 2, …, 10 in the infinite net, counting images;
TD10 is the running total) and the point/vertex symbol. Compare with the RCSR table for
nets of that connectivity:
dia(4-c: 4, 12, 24, 42, 64, …),pcu(6-c: 6, 18, 38, 66, 102, …),bcu(8-c: 8, 26, 56, 98, 152, …),srs,ths,qtz,lon,acs,nbo,cdsand so on. A coordination sequence that matches no RCSR entry means the graph is wrong more often than the net is new; go back to steps 3–4. - Distances and averages. An internodal distance is the centroid-to-centroid length of an edge, computed with the correct image vector. Report the mean over all distinct edges (per node class if the task asks), in the units and rounding the task states, and print the count of edges that entered the mean.
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.
- 20d ago First seen · 102 lines · 128 tokens per session scan A 0dd613bfe0cc
framework-topology is a skill published in the GitHub repository synthetic-sciences/openscience (4,025 stars, last pushed today), licensed Apache-2.0. It adds 128 tokens to every session and 1,682 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 2026-09-27.
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.