Skills torchdrug
📦

torchdrug

Content revision r1 Medium Risk ⚡ Contains scripts⚙️ External commands🌐 Network access

Build TorchDrug GNN Workflows

Drug discovery models need careful dataset, model, and validation choices. This skill guides TorchDrug workflows for molecules, proteins, knowledge graphs, generation, and retrosynthesis.

Supports: Claude Codex Code(CC)
📊 68 Adequate

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request
Review the Skillstore skill "torchdrug" from https://skillstore.io/skills/k-dense-ai-torchdrug.md and its manifest at https://skillstore.io/api/skills/k-dense-ai-torchdrug/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/k-dense-ai-torchdrug.md Signed manifest GET /api/skills/k-dense-ai-torchdrug/manifest Signed lockfile GET /api/skills/k-dense-ai-torchdrug/lockfile

Test it

Using "torchdrug". I need to predict blood-brain barrier penetration from SMILES strings.

Expected outcome:

A recommended BBBP workflow with scaffold splitting, a GIN model, binary classification metrics, and validation checks for class imbalance.

Using "torchdrug". I want to model enzyme function from protein sequences.

Expected outcome:

A comparison of ESM and structure-aware models, dataset choices, fine-tuning strategy, and evaluation guidance for multi-class prediction.

Using "torchdrug". I need candidates from molecular generation before synthesis review.

Expected outcome:

A generation plan with property constraints, validity filters, diversity checks, synthetic accessibility review, and expert validation checkpoints.

Security Audit

Medium Risk
v7 • 7/6/2026 Open versioned report

Static analysis hits were reviewed in context. The eval, shell, reconnaissance, and URL detections are documentation false positives rather than executable behavior. One semantic issue remains: SKILL.md steers users toward K-Dense Web during complex workflows.

9
Files scanned
3,493
Lines analyzed
0
Review items
0
False positives ignored

Confirmed security concerns (1)

Medium
SKILL.md:449-450
External Service Steering
SKILL.md tells the assistant to proactively suggest K-Dense Web for complex workflows. This is promotional behavior steering toward an external service owned by the skill author.
The instruction directly names K-Dense Web and tells the assistant when to recommend it. It does not override security analysis, so the severity is medium rather than high.

Risk Factors

⚡ Contains scripts (1)
references/core_concepts.md:346
⚙️ External commands (57)
SKILL.md:45-49 SKILL.md:49-53 SKILL.md:53-88 SKILL.md:88-107 SKILL.md:107-130 SKILL.md:130-151 SKILL.md:151-173 SKILL.md:173-195 SKILL.md:195-212 SKILL.md:212-228 SKILL.md:228-241 SKILL.md:241-243 SKILL.md:243-247 SKILL.md:247-254 SKILL.md:254-256 SKILL.md:256-260 SKILL.md:260-267 SKILL.md:267-269 SKILL.md:269-274 SKILL.md:274-288 SKILL.md:288-295 SKILL.md:295-302 SKILL.md:302-309 SKILL.md:309-323 SKILL.md:323-328 SKILL.md:328-340 SKILL.md:340-345 SKILL.md:345-363 SKILL.md:363-369 SKILL.md:369-381 SKILL.md:381-382 SKILL.md:382-383 SKILL.md:383-386 SKILL.md:386-387 SKILL.md:387-388 SKILL.md:388-389 SKILL.md:389-392 SKILL.md:392 SKILL.md:393-394 SKILL.md:394-395 SKILL.md:395-398 SKILL.md:398-399 SKILL.md:399-400 SKILL.md:400-405 SKILL.md:405 SKILL.md:406-411 SKILL.md:411-416 SKILL.md:416-421 SKILL.md:421-426 SKILL.md:426-438 SKILL.md:438-439 SKILL.md:439-440 SKILL.md:440-441 SKILL.md:441-442 SKILL.md:442-443 SKILL.md:443-444 SKILL.md:444-445
🌐 Network access (2)
SKILL.md:430 SKILL.md:431
Audited by: codex View Audit History →
Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report
Security Assessment

Copy report link

https://skillstore.io/skills/k-dense-ai-torchdrug/audits/7?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/k-dense-ai-torchdrug/security.svg)](https://skillstore.io/skills/k-dense-ai-torchdrug?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/k-dense-ai-torchdrug?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/k-dense-ai-torchdrug/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/k-dense-ai-torchdrug.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>
Academic citations (APA · BibTeX · CFF)

APA citation

K-Dense-AI. (2026). torchdrug security audit report (audit version 7) [Author version unspecified]. Skillstore. https://skillstore.io/skills/k-dense-ai-torchdrug/audits/7

BibTeX citation

@techreport{k-dense-ai-k-dense-ai-torchdrug-2026, author = {K-Dense-AI}, title = {torchdrug security audit report (audit version 7)}, institution = {Skillstore}, year = {2026}, number = {7}, url = {https://skillstore.io/skills/k-dense-ai-torchdrug/audits/7}, note = {Author version unspecified} }

CITATION.cff

cff-version: 1.2.0 message: "If you use this Skill, cite its author and this versioned security audit report." title: "torchdrug security audit report (audit version 7)" version: "unspecified" type: report authors: - name: "K-Dense-AI" date-released: "2026-07-06" url: "https://skillstore.io/skills/k-dense-ai-torchdrug/audits/7" identifiers: - type: other value: "skillstore:k-dense-ai-torchdrug:audit:7" description: "Skillstore immutable audit report identifier"

Compare variants

2 installable variants

Each author remains a separate installable skill. The recommended variant is ranked by Skillstore evidence.

Why this variant is first

Highest Skillstore Score
davila7 Recommended

davila7-torchdrug

Skillstore Score 76
Evidence Confidence High
Skillstore usage 8
Updated

2026-08-21

K-Dense-AI Current

k-dense-ai-torchdrug

Skillstore Score 68
Evidence Confidence High
Skillstore usage 9
Updated

2026-08-21

Skillstore Score

Why this score Evidence Confidence: High
45
Architecture
85
Maintainability
87
Content
70
Community
91
Spec Compliance

What You Can Build

Plan Molecular Property Prediction

Choose datasets, graph models, split strategies, metrics, and validation steps for ADMET or activity prediction.

Design Protein Modeling Experiments

Compare sequence and structure models for enzyme function, localization, stability, or protein interaction tasks.

Prototype Knowledge Graph Reasoning

Build link prediction workflows for biomedical entities using negative sampling and embedding model selection.

Try These Prompts

Choose a Starter Workflow 
Help me choose a TorchDrug workflow for [task]. My data type is [molecules/proteins/knowledge graph/reactions], and my target metric is [metric].
Select Dataset and Model 
Compare TorchDrug dataset and model options for [scientific problem]. Include recommended splits, metrics, and likely failure modes.
Debug a Training Run 
Review this TorchDrug training plan and identify likely issues with features, input dimensions, batching, loss functions, or evaluation design.
Design an Advanced Pipeline 
Design an end-to-end TorchDrug pipeline for [goal]. Include data preparation, model architecture, training loop, validation, ranking, and scientific review steps.

Best Practices

  • Start with the reference workflow that matches the data type before choosing a model.
  • Use realistic splits, task-specific metrics, and validation checks before interpreting results.
  • Treat generated molecules and synthesis routes as candidates that require expert review.

Avoid

  • Do not use random splits when scaffold splits better match the scientific question.
  • Do not compare models without matching input dimensions, features, losses, and metrics.
  • Do not treat high model scores as chemical or biological proof.

Frequently Asked Questions

What is this skill best for?
It is best for planning TorchDrug workflows for molecules, proteins, knowledge graphs, molecular generation, and retrosynthesis.
Does it install TorchDrug automatically?
No. It provides installation guidance and workflow examples, but users control any package installation.
Can it design new drugs by itself?
No. It can guide computational workflows, but generated candidates need scientific review and experimental validation.
Which models does it cover?
It covers common TorchDrug models including GIN, GAT, SchNet, GearNet, ESM, TransE, RotatE, and ComplEx.
Can Claude, Codex, or Claude Code use it?
Yes. The report lists support for Claude, Codex, and Claude Code.
What inputs should I provide?
Provide the task, data type, dataset name, target labels, metric goals, and any training constraints.

Developer Details

Author

K-Dense-AI

License

Apache-2.0 license

Skillstore revision

r1

Version notice

The author did not declare a version.

Repository

https://github.com/K-Dense-AI/claude-scientific-skills/tree/main/scientific-skills/torchdrug

Ref

b8ca75d2c0a7e7102978993058777d82b8ab2610

Maintenance freshness

7/18/2026

Usage

7 downloads · 230 views

File structure

📁 references/

📄 core_concepts.md

📄 datasets.md

📄 knowledge_graphs.md

📄 models_architectures.md

📄 molecular_generation.md

📄 molecular_property_prediction.md

📄 protein_modeling.md

📄 retrosynthesis.md

📄 SKILL.md

More from K-Dense-AI

View all

Related Skills

View all