SPYfighting/my-academic-skills

**Prompt and data release** — *Claude protein binder design (data release v1.0)*, Anthropic, on HuggingFace Datasets, licensed **CC-BY-4.0**: <https://huggingface.co/datasets/Anthropic/claude-protein-binder-design>

<https://huggingface.co/datasets/Anthropic/claude-protein-binder-design>

**Technical report** — *Autonomous de novo protein binder design with Claude*, Claude Science and Amir Shanehsazzadeh, Anthropic, 18 August 2026: <https://www-cdn.anthropic.com/30bf50e22a01388bb29bf077ee3f244531594b7a.pdf>

<https://www-cdn.anthropic.com/30bf50e22a01388bb29bf077ee3f244531594b7a.pdf>

**Announcement** — *How Claude is accelerating protein design and analytical chemistry*, Anthropic: <https://www.anthropic.com/research/Claude-accelerates-protein-design>

<https://www.anthropic.com/research/Claude-accelerates-protein-design>

**Adaptyv Bio protein design skills** — MIT. <https://github.com/adaptyvbio/protein-design-skills>

<https://github.com/adaptyvbio/protein-design-skills>

**Claude Science skill set**, mirrored in AcademicForge — MIT. <https://github.com/HughYau/AcademicForge/tree/site-first/skills/claude-science>

<https://github.com/HughYau/AcademicForge/tree/site-first/skills/claude-science>

**Anthropic skills** (specification and template) — <https://github.com/anthropics/skills>

<https://github.com/anthropics/skills>

#!/usr/bin/env python3

#!/usr/bin/env python3

#!/usr/bin/env python3

#!/usr/bin/env python3

#!/usr/bin/env python3

#!/usr/bin/env python3

`multi_target_binder_design_prompt.md` is an unmodified copy of Anthropic's multi-target campaign prompt, downloaded from its [official Hugging Face data release](https://huggingface.co/datasets/Anthropic/claude-protein-binder-design/blob/m

multi-target campaign prompt, downloaded from its [official Hugging Face data release](https://huggingface.co/datasets/Anthropic/claude-protein-binder-design/blob/main/prompts/prompts/multi_target_binder_design_prompt.md) on 2026-09-30. SHA

release](https://huggingface.co/datasets/Anthropic/claude-protein-binder-design/blob/main/prompts/prompts/multi_target_binder_design_prompt.md)

1. [Adaptyv blog: BenchBB benchmark and community paper](https://www.adaptyvbio.com/blog/benchbb/)

2. [Adaptyv blog: Protein Optimization 101](https://www.adaptyvbio.com/blog/po101)

3. [Adaptyv blog: Protein Optimization 102](https://www.adaptyvbio.com/blog/po102)

4. [Adaptyv blog: Protein Optimization 103](https://www.adaptyvbio.com/blog/po103)

5. [Adaptyv blog: Protein Design Competition: has binder design been solved?](https://www.adaptyvbio.com/blog/po104/)

6. [Adaptyv blog: MBP case study: modular protein sensors with BindCraft](https://www.adaptyvbio.com/blog/mbp)

7. [bioRxiv: Crowdsourced protein design, Adaptyv EGFR competition](https://www.biorxiv.org/content/10.1101/2025.04.17.648362v2)

8. [ProteinBase: Adaptyv EGFR binder competition, round 1](https://proteinbase.com/collections/adaptyv-egfr-competition-round-1)

9. [ProteinBase: Adaptyv EGFR binder competition, round 2](https://proteinbase.com/collections/adaptyv-egfr-competition-round-2)

10. [Adaptyv blog: Agents vs Humans: can LLMs design proteins?](https://www.adaptyvbio.com/blog/agents-vs-humans)

11. [ProteinBase: Adaptyv × MUNI hackathon: AI agents vs humans](https://proteinbase.com/collections/adaptyv-x-muni-hackathon-ai-agents-vs-humans)

12. [ProteinBase: BioArena × Adaptyv TREM2 binder competition](https://proteinbase.com/competitions/bioarena-adaptyv-trem2)

13. [ProteinBase: EPFL × Adaptyv Myostatin (GDF-8) binder competition](https://proteinbase.com/competitions/epfl-adaptyv-myostatin)

14. [GEM Bio: GEM × Adaptyv 2026 RBX1 binder design competition](https://www.gembio.ai/gemadaptyv2026)

15. [ProteinBase: GEM × Adaptyv RBX1 competition: results](https://proteinbase.com/collections/gem-x-adaptyv-rbx1-binder-design-competition-results)

16. [ProteinBase: GEM × Adaptyv RBX1 competition: all submissions](https://proteinbase.com/collections/gem-adaptyv-rbx1-competition-all-submissions)

17. [Adaptyv blog: Nipah-G binder competition: submissions overview](https://www.adaptyvbio.com/blog/nipah-submissions/)

18. [ProteinBase: Adaptyv Nipah-G binder competition](https://proteinbase.com/competitions/adaptyv-nipah-competition)

19. [ProteinBase: Nipah-G binder competition: results](https://proteinbase.com/collections/nipah-binder-competition-results)

20. [ProteinBase: Nipah-G binder competition: all submissions](https://proteinbase.com/collections/nipah-binder-competition-all-submissions)

21. [ProteinBase: RFdiffusion re-validation](https://proteinbase.com/collections/rfdiffusion-re-validation)

22. [ProteinBase: BindCraft v1 re-validation](https://proteinbase.com/collections/bindcraft1-revalidation)

23. [arXiv: AlphaProteo technical report](https://arxiv.org/abs/2409.08022)

24. [DeepMind blog: AlphaProteo](https://deepmind.google/blog/alphaproteo-generates-novel-proteins-for-biology-and-health-research/)

25. [Nature 2022: Cao et al., Design of protein-binding proteins from the target structure alone](https://www.nature.com/articles/s41586-022-04654-9)

26. [Nature 2023: Watson et al., RFdiffusion](https://www.nature.com/articles/s41586-023-06415-8)

27. [Nature 2025: Pacesa et al., BindCraft](https://www.nature.com/articles/s41586-025-09429-6)

28. [BoltzGen](https://www.biorxiv.org/content/10.1101/2025.11.20.689494v1)

29. [ProteinBase: BoltzGen release designs](https://proteinbase.com/collections/boltzgen-release)

1. ESMFold2 / ESMFold2-Fast: weights at HuggingFace biohub/ESMFold2 and biohub/ESMFold2-Fast (ungated, MIT); install the esm package from [github.com/Biohub/esm](https://github.com/Biohub/esm) on Python 3.12 (the PyPI "esm" package is a dif

2. Protenix v2: install from [github.com/bytedance/Protenix](https://github.com/bytedance/Protenix) (pip install protenix==2.0.0, CUDA 12.4 devel base); stage the v2 checkpoint and CCD cache from [huggingface.co/TMF001/pxdesign-weights](htt

3. AlphaFold3 code + OpenFold3 weights: AF3 JAX engine running OF3 open weights, fully Apache-2.0. Build per [github.com/sokrypton/alphafold3](https://github.com/sokrypton/alphafold3) (OF3_AF3_PORTING_NOTES.md), port of3-p2-155k.pt from [hu

For every model or tool the prompt names, the install + invocation recipe must be derived from the project's GitHub README / HuggingFace card / paper, unless a skill exists (in which case you should still review the project's GitHub README 

You must utilize *in silico* optimization by applying test-time compute to raise scores on the most promising designs, using approaches such as [partial diffusion with sequence redesign](https://www.nature.com/articles/s41586-023-06953-1); 

Compute: You are to use Modal for access to GPUs to run design and filtering models. The governor caps concurrent GPU instances. Modal bills GPU + CPU + memory per second, separately. Published per-second GPU rates ([modal.com/pricing](http

[Credential-related source omitted]

This skill adapts and restructures public MIT-licensed material from [yrui-cmd/cell-lct](https://github.com/yrui-cmd/cell-lct/tree/4af98d8057f48a944141f9122da7aba83d544d28). The reviewed upstream commit was `4af98d8057f48a944141f9122da7aba8

[VTracer 1.0.0-alpha.4](https://github.com/visioncortex/vtracer/releases/tag/1.0.0-alpha.4) is an optional, separately installed dependency under its own license. Its installer downloads only a pinned official release asset and verifies SHA

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 640 320" width="640" height="320">

Conventions checked on 2026-09-08: [Codex](https://developers.openai.com/codex/skills), [Claude Code](https://code.claude.com/docs/en/skills), [OpenCode](https://opencode.ai/docs/skills/). Image understanding, sandbox access, and native-app

Adobe documented an Illustrator 25.1 AppleScript `do javascript` regression and its fix in 25.2. Update affected installations rather than promising to bypass an application defect. [Adobe version note](https://blog.developer.adobe.com/en/p

Executable file

#!/usr/bin/env python3

Executable file

Executable file

Executable file

Executable file

Executable file

Executable file

NS = 'http://www.w3.org/2000/svg'

Executable file

Executable file

[Credential-related source omitted]

Executable file

#!/usr/bin/env python3

target=destination/('vtracer.exe' if key[0]=='Windows' else 'vtracer')     if target.exists(): raise FileExistsError(f'{target} already exists; refusing silent replacement')     url=f'https://github.com/visioncortex/vtracer/releases/downloa

if target.exists(): raise FileExistsError(f'{target} already exists; refusing silent replacement')     url=f'https://github.com/visioncortex/vtracer/releases/download/{VERSION}/{asset}'     request=urllib.request.Request(url,headers={'User-

url=f'https://github.com/visioncortex/vtracer/releases/download/{VERSION}/{asset}'

request=urllib.request.Request(url,headers={'User-Agent':'Scientific-Illustrator-Local-Installer/0.1'})

with urllib.request.urlopen(request,timeout=60) as response:

#!/usr/bin/env python3

for index, body in enumerate([             '<image href="https://invalid.example/image.png" width="10" height="10"/>',

'<image href="https://invalid.example/image.png" width="10" height="10"/>',

]):             invalid = temp / ('invalid-%s.svg' % index)             invalid.write_text('<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20">' + body + '</svg>', encoding='utf-8')

invalid = temp / ('invalid-%s.svg' % index)             invalid.write_text('<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20">' + body + '</svg>', encoding='utf-8')             rejected(lambda: compile_svg(invalid))

invalid.write_text('<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20">' + body + '</svg>', encoding='utf-8')

loop = temp / 'loop.svg'         loop.write_text('<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20"><path d="M0 0 C10 0 0 10 0 0Z"/></svg>', encoding='utf-8')

loop.write_text('<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20"><path d="M0 0 C10 0 0 10 0 0Z"/></svg>', encoding='utf-8')

[Credential-related source omitted]

#!/usr/bin/env python3

This skill adapts SVG parsing, tracing, text-manifest, and local preview modules from the user's [scientific-illustrator-local](https://github.com/SPYfighting/my-academic-skills/tree/98e4152/scientific-illustrator-local) skill at commit `98

The PowerPoint DrawingML renderer, application workflow, and figure-specific validation were added here. It is a local tool; it does not include Adobe software, a cloud vectorization service, model weights, or fonts. `VTracer` 1.0.0a4 is an

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 120 80">

For Windows x86-64, the pinned releases of [Pillow](https://pypi.org/project/pillow/12.3.0/), [NumPy](https://pypi.org/project/numpy/2.3.5/), and [OpenCV headless](https://pypi.org/project/opencv-python-headless/4.13.0.92/) list Windows whe

The bundled blank template and renderer use base PresentationML and DrawingML shapes and live text. They do not require Office 2010+ extensions, macros, linked resources, embedded pictures, SmartArt, or visual effects. Microsoft lists `.ppt

#!/usr/bin/env python3

def inspect(pptx: Path) -> dict:     ns = {'p': 'http://schemas.openxmlformats.org/presentationml/2006/main',

ns = {'p': 'http://schemas.openxmlformats.org/presentationml/2006/main',

'a': 'http://schemas.openxmlformats.org/drawingml/2006/main'}

Executable file

Executable file

Executable file

P = 'http://schemas.openxmlformats.org/presentationml/2006/main'

A = 'http://schemas.openxmlformats.org/drawingml/2006/main'

content.text = text     if text != text.strip():         content.set('{http://www.w3.org/XML/1998/namespace}space', 'preserve')

if text != text.strip():         content.set('{http://www.w3.org/XML/1998/namespace}space', 'preserve')     sub(paragraph, A, 'endParaRPr', lang='en-US')

content.set('{http://www.w3.org/XML/1998/namespace}space', 'preserve')

versioned = re.compile(rb'http://schemas\.microsoft\.com/office/[^"<> ]+/20\d\d/')

Executable file

Executable file

NS = 'http://www.w3.org/2000/svg'

Executable file

Executable file

[Credential-related source omitted]

#!/usr/bin/env python3

bad_svg = temp / 'embedded.svg'         bad_svg.write_text('<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20">'

bad_svg.write_text('<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 20 20">'

'<image href="https://invalid.example/x.png" width="10" height="10"/>'

if entry.filename == 'ppt/slides/slide1.xml':                     root = ET.fromstring(data)                     ET.SubElement(root, '{http://schemas.microsoft.com/office/powerpoint/2010/main}creationId', val='1')

root = ET.fromstring(data)                     ET.SubElement(root, '{http://schemas.microsoft.com/office/powerpoint/2010/main}creationId', val='1')                     data = ET.tostring(root, encoding='utf-8', xml_declaration=True)

ET.SubElement(root, '{http://schemas.microsoft.com/office/powerpoint/2010/main}creationId', val='1')