torchdrug
PyTorch-native graph neural networks for molecules and proteins. Use when building custom GNN architectures for drug discovery, protein modeling, or knowledge graph reasoning. Best for custom model development, protein property prediction, retrosynthesis. For pre-trained models and diverse featurizers use deepchem; for benchmark datasets use pytdc.
Install with Tessl CLI
npx tessl i github:K-Dense-AI/claude-scientific-skills --skill torchdrugOverall
score
86%
Does it follow best practices?
If you maintain this skill, you can automatically optimize it using the tessl CLI to improve its score:
npx tessl skill review --optimize ./path/to/skillValidation for skill structure
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is an excellent skill description that clearly defines a specific technical domain (PyTorch GNNs for molecular/protein modeling), provides explicit trigger conditions, and thoughtfully differentiates itself from related skills. The inclusion of when to use alternative tools (deepchem, pytdc) is particularly valuable for skill selection in a multi-skill environment.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: 'building custom GNN architectures', 'drug discovery', 'protein modeling', 'knowledge graph reasoning', 'protein property prediction', 'retrosynthesis'. These are concrete, domain-specific capabilities. | 3 / 3 |
Completeness | Clearly answers both what ('PyTorch-native graph neural networks for molecules and proteins') and when ('Use when building custom GNN architectures for drug discovery, protein modeling, or knowledge graph reasoning'). Also includes explicit guidance on when NOT to use it (directing to deepchem or pytdc for other use cases). | 3 / 3 |
Trigger Term Quality | Excellent coverage of natural terms users would say: 'PyTorch', 'graph neural networks', 'GNN', 'molecules', 'proteins', 'drug discovery', 'protein modeling', 'knowledge graph', 'retrosynthesis'. Also includes differentiation terms like 'deepchem' and 'pytdc'. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive with clear niche (PyTorch GNNs for molecular/protein work) and explicit differentiation from related tools (deepchem for pre-trained models, pytdc for benchmarks). This boundary-setting significantly reduces conflict risk. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
73%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a well-structured skill with excellent progressive disclosure and actionable code examples. The main weaknesses are moderate verbosity (especially the 'When to Use' section and promotional content) and workflows that lack explicit validation checkpoints for complex multi-step operations like molecular generation and retrosynthesis.
Suggestions
Remove or significantly condense the 'When to Use This Skill' section - Claude can infer appropriate use cases from the capabilities described
Remove the promotional K-Dense Web section at the end as it's not relevant to teaching the skill
Add explicit validation steps to workflows, especially for molecular generation (e.g., 'Validate generated SMILES with RDKit before property prediction') and retrosynthesis (e.g., 'Verify reaction feasibility before proceeding to next step')
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill contains some unnecessary verbosity, particularly in the 'When to Use This Skill' section which lists obvious use cases Claude would infer. The overview repeats information, and the promotional K-Dense section at the end is irrelevant to the skill's purpose. | 2 / 3 |
Actionability | Provides fully executable code examples including installation, dataset loading, model creation, training loops, and integration patterns with RDKit, AlphaFold, and PyTorch Lightning. Code is copy-paste ready with proper imports. | 3 / 3 |
Workflow Clarity | Workflows are listed with clear steps but lack explicit validation checkpoints. For example, the molecular generation workflow doesn't include validation steps between generation and filtering, and retrosynthesis planning lacks error handling for failed predictions. | 2 / 3 |
Progressive Disclosure | Excellent structure with clear overview sections pointing to one-level-deep reference files. Navigation is well-signaled with a Quick Reference Cheat Sheet and clear file references for each capability area. | 3 / 3 |
Total | 10 / 12 Passed |
Validation
94%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 15 / 16 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
metadata_version | 'metadata.version' is missing | Warning |
Total | 15 / 16 Passed | |
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.