torch-geometric

Graph Neural Networks (PyG). Node/graph classification, link prediction, GCN, GAT, GraphSAGE, heterogeneous graphs, molecular property prediction, for geometric deep learning.

1.19x

Quality

77%

Does it follow best practices?

Impact

98%

1.19x

Average score across 3 eval scenarios

Securityby

Advisory

Suggest reviewing before use

Optimize this skill with Tessl

npx tessl skill review --optimize ./scientific-skills/torch_geometric/SKILL.md

Evaluation results

100%

Graph Convolution Layer from Scratch

Custom MessagePassing layer implementation

Criteria

Without context

With context

MessagePassing base class

100%

aggr parameter set

100%

add_self_loops utility

100%

degree utility

100%

Symmetric normalization

100%

propagate() called

100%

_j suffix in message()

100%

edge_index dtype

100%

GPU device handling

100%

Implementation runs

100%

95%

38%

Scalable Node Classification for Large Graphs

Large-scale NeighborLoader training pipeline

Criteria

Without context

With context

NeighborLoader used

83%

num_neighbors hops

100%

Seed node loss only

100%

input_nodes mask

100%

SAGEConv or scalable layer

100%

GPU device handling

87%

100%

DataLoader import

100%

Directed subgraph note

57%

Training report

100%

Runs end-to-end

100%

10%

Molecular Graph Dataset Pipeline

Custom InMemoryDataset with transforms

Criteria

Without context

With context

InMemoryDataset inheritance

100%

self.load() in __init__

100%

raw_file_names property

100%

processed_file_names property

100%

pre_filter check in process()

100%

pre_transform check in process()

100%

self.save() in process()

100%

Compose transforms used

100%

edge_index dtype

100%

Dataset instantiated and used

100%

Repository: K-Dense-AI/claude-scientific-skills
Commit: b271271

Evaluated: about 2 months ago
Agent: Claude Code
Model: Claude Sonnet 4.6

Table of Contents

Graph Convolution Layer from Scratch Scalable Node Classification for Large Graphs Molecular Graph Dataset Pipeline

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.