Darshil patel

Member of Technical Staff @k_dense_ai

[K-Dense @k_dense_ai]

Thank you to all for sharing and using our scientific-agent-skills repo, which hit 40k stars this weekend!

[Darshil patel @Darshil2710]

40,000 stars. ⭐️

Scientific Agent Skills is now the open standard for turning any AI agent into a working scientist.

165+ skills. 100+ databases. Built in the open.

https://t.co/F4QlK8bkzx

@k_dense_ai

[liam. @liambraus]

163 SKILLS PARA CONVERTIR TU AGENTE EN CIENTÍFICO

Un repo open source acaba de convertir cualquier agente de ia en un científico con acceso a genómica, química computacional y docking molecular sin escribir una sola integración.

Se llama Scientific Agent Skills.

Ya lo usan más de 190.000 científicos, y suma 40.400 estrellas en GitHub.

Así funciona:

1. Instalas con un solo comando: npx skills add
2. Tu agente gana acceso a 163 skills y 100+ bases de datos científicas (pubchem, chembl, uniprot, clinicaltrialsgov...)
3. Cada skill trae documentación, tests y ejemplos reales listos para usar

Compatible con Claude Code, Cursor, Codex y Antigravity.

Guárdate esta herramiena. Repo abajo

[Darshil patel @Darshil2710]

@FareaNFts try : https://t.co/F4QlK8bkzx

[divyansh tiwari @DivyanshT91162]

This GitHub repo turns your AI agent into a scientific research assistant.

Meet Scientific Agent Skills by K-Dense.

It gives AI agents 163 ready-to-use skills for complex research workflows.

What’s inside?

→ 100+ scientific databases
→ 70+ optimized Python package skills
→ 9 scientific integrations
→ Literature research & evidence synthesis
→ Bioinformatics & genomics
→ Drug discovery
→ Protein engineering
→ Medical imaging
→ ML & data analysis
→ Physics, astronomy & geospatial research

It has dedicated workflows for tools like:

RDKit, Scanpy, BioPython, scikit-learn, Qiskit, GeoPandas, PyTorch Lightning, OpenMM and more.

And it works with Claude Code, Codex, Cursor, Gemini CLI, Google Antigravity and other Agent Skills-compatible tools.

There’s also K-Dense BYOK, an open-source AI co-scientist that runs locally with your own API keys.

The crazy part?

You’re not just giving your AI more knowledge.

You’re giving it specialized skills to actually perform research workflows.

REPOO👇

[Tom Dörr @tom_doerr]

Open-source scientific agent skills library featuring 163 workflows for cancer genomics, drug discovery, and clinical research.

https://t.co/rqBPybOsgH

[K-Dense @k_dense_ai]

Introducing K-Bench 01. A new private benchmark composed of 178 end-to-end scientific tasks.

Read the pre-print: https://t.co/ucbOvV1jWf

[Timothy Kassis @TimothyKassis]

Our Scientific Agent Skills is now one of the most popular open source projects that have ever existed just surpassing 37k+ GitHub stars. https://t.co/cFLvrfkhQh

[Darshil patel @Darshil2710]

We just added a highly requested feature to K-Dense Web: Data Connectors.

Connect your private Amazon S3 or Google Cloud Storage and import files directly into a research session - no more downloading and re-uploading.

Connections are read-only and encrypted. We never get write access to your buckets.

Learn how: https://t.co/J94qwK1qsu

@k_dense_ai

[Timothy Kassis @TimothyKassis]

Learn how to use K-Dense BYOK (a better alternative to Claude Science) though this 45 min tutorial with @ambrueckner and myself: https://t.co/rjA3NzN0RT

[Darshil patel @Darshil2710]

Proud to share K-Bench 01 from @k_dense_ai.

We evaluated nine frontier models on 178 real scientific agent requests drawn from live user traffic, resulting in 1,602 controlled runs. No model cleared the acceptability threshold under all three independent judges. Overclaiming was the leading failure mode at 31.4%, and the top-ranked model overall was not the one that uniquely solved the hardest tasks.

Full preprint:
https://t.co/vX2rYk1X1w

[Timothy Kassis @TimothyKassis]

New paper from @k_dense_ai : K-Bench 01.

No model cleared our acceptable bar under all three judges. Overclaiming was the top failure at 31.4%. The best model on the leaderboard was not the one that uniquely solved the hardest tasks. 178 real user requests, 9 models, 1,602 runs.

Thanks to @ambrueckner, @Darshil2710 and @yhhonx for all the work that made this happen.

[Darshil patel @Darshil2710]

Science deserves tools that put researchers first.

Next Tuesday, August 25 at 2 PM PST, we’re introducing K-Dense BYOK: an open-source platform that lets you run a powerful AI research assistant locally - or with your own keys - so your data and costs remain fully under your control.

Accessible. Practical. No specialized technical knowledge needed.

We’d love to have you with us.

Register here: https://t.co/ZCMSij5The

[K-Dense @k_dense_ai]

Happening next week - live!

On August 25 at 2pm PST we'll introduce K-Dense BYOK (bring-your-own keys) an open source platform for science that you can run locally or with your own API keys, keeping you in control of cost and data. ​

The session is designed to be accessible and practical, so come ready to engage and learn. No prior experience with the tool or specialized technical knowledge is required.

Registration required: https://t.co/ua2Slo02qa

[K-Dense @k_dense_ai]

New tutorial from our team on how to use @MulticaAI with Scientific Agent Skills as a team of AI co-scientists. Read this tutorial by our CTO, @TimothyKassis, and let us know what you think: https://t.co/wA264JUMo3

[Darshil patel @Darshil2710]

Introducing our first focused plugin: Small-Molecule & Protein Therapeutics

A carefully curated collection of agent skills for drug discovery, including cheminformatics, molecular machine learning, docking & dynamics, protein design platforms, and target-discovery knowledge graphs.

Install the complete bundle as a ready-to-use plugin in Claude Code, Cursor, Codex, Google Antigravity, or any compatible agentic harness.

Start accelerating therapeutics research today.

[K-Dense @k_dense_ai]

We just shipped our first agent skills plugin for small molecule and protein therapeutics. This represents a curated subset of our larger Scientific Agent Skills repo (33k+ ⭐ on GitHub).

Try it today in Claude Code, Cursor, Codex, Google Antigravity, and more.

https://t.co/bfRcHZm7DR

[K-Dense @k_dense_ai]

We just shipped our first agent skills plugin for small molecule and protein therapeutics. This represents a curated subset of our larger Scientific Agent Skills repo (33k+ ⭐ on GitHub).

Try it today in Claude Code, Cursor, Codex, Google Antigravity, and more.

https://t.co/bfRcHZm7DR

[Darshil patel @Darshil2710]

Introducing lab-hardware-cad - an elegant new addition to our Scientific Agent Skills.

Empower your AI agents to design fabrication-ready laboratory hardware with standards-compliant accuracy and refined parametric control.

Explore the full benchmarking study:
https://t.co/fShJqCw366

[K-Dense @k_dense_ai]

Introducing the lab-hardware-cad skill! Now it's easier for your AI agents to design lab hardware such as microfluidic devices, plate holders...etc. Full writeup and benchmarking: https://t.co/ABwz1Yz0NB

[K-Dense @k_dense_ai]

Follow the K-Dense agent's progress during phase 2 of the "MecCog agent collaboration to map Alzheimer's research" https://t.co/pEoWdhJ7p7