torchdrug-english

PyTorch-native Graph Neural Network framework for molecules and proteins. Suitable for building custom GNN architectures for drug discovery, protein modeling, or knowledge graph reasoning. Best for custom model development, protein property prediction, and retrosynthesis. If you need pretrained models and diverse feature extractors, use deepchem; if you need benchmark datasets, use pytdc.

Quality

77%

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Passed

No known issues

Fix and improve this skill with Tessl

tessl review fix ./scientific-skills/Data Analysis/TorchDrug-English/SKILL.md

Quality

Content

72%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The body is highly actionable with executable code and specific API references, and it uses clean one-level-deep progressive disclosure into the references/ bundle. Its weaknesses are verbosity from triple-repeated description text and generic templated boilerplate, plus workflows that lack explicit validation checkpoints.

Suggestions

Remove the verbatim description repetition in 'When to Use' and 'Key Features' and delete the generic templated bullets ('a data analytics task needs a packaged method', 'the user expects a concrete deliverable') that are not specific to TorchDrug.

Add explicit validation/verification checkpoints to the Common Workflows (e.g., confirm scaffold split integrity, check loss convergence, verify AUROC/AUPRC against expected ranges before reporting).

Either reference core_concepts.md from the body or remove it from references/ so every bundle file is reachable via signaled navigation.

Dimension	Reasoning	Score
Conciseness	The frontmatter description is repeated verbatim three times (frontmatter, 'When to Use', 'Key Features') and the 'When to Use'/'Implementation Details' sections are generic templated boilerplate ('a data analytics task needs a packaged method', 'validate the request, choose the packaged workflow'); the technical core is lean but padded by this filler, so it is mostly efficient with notable unnecessary content.	2 / 3
Actionability	The Quick Example is fully executable copy-paste Python (imports, datasets.BBBP, models.GIN, tasks.PropertyPrediction, training loop) and the capability/workflow sections name specific datasets, models, and task classes.	3 / 3
Workflow Clarity	The three Common Workflows list clearly sequenced steps (Load → Choose model → Define task → Train → Evaluate) with navigation pointers, but no validation checkpoints or error-recovery feedback loops are present, matching the 'steps listed but validation gaps' anchor.	2 / 3
Progressive Disclosure	SKILL.md is an overview that splits detail into one-level-deep, clearly signaled reference files ('Reference: See molecular_property_prediction.md') with section-level navigation pointers; all but one bundle file (core_concepts.md) are referenced, matching the clear-overview-with-well-signaled-references anchor.	3 / 3
	Total	10 / 12 Passed

Description

82%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A specific, distinctive description with good domain trigger terms and explicit conflict-avoidance guidance against deepchem and pytdc. Its main weakness is the absence of an explicit 'Use when…' trigger clause, leaving the 'when' implied rather than stated.

Suggestions

Add an explicit 'Use when …' clause naming user-facing triggers (e.g., 'Use when the user wants to train or fine-tune GNNs on molecular graphs, protein sequences/structures, or biomedical knowledge graphs').

Reduce repetition between 'Suitable for…' and 'Best for…' by merging them into a single concrete trigger sentence.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete capabilities — 'building custom GNN architectures for drug discovery, protein modeling, or knowledge graph reasoning' and 'protein property prediction, and retrosynthesis' — matching the anchor for several specific concrete actions.	3 / 3
Completeness	The 'what' is strong but there is no explicit 'Use when…' trigger clause; the 'when' is only implied via 'Suitable for…/Best for…', so per the guideline completeness is capped at 2 rather than reaching the explicit-trigger anchor of 3.	2 / 3
Trigger Term Quality	Natural domain keywords a specialist user would say are well covered: 'drug discovery', 'protein modeling', 'knowledge graph reasoning', 'retrosynthesis', 'GNN'/'Graph Neural Network', 'molecules and proteins'.	3 / 3
Distinctiveness Conflict Risk	It explicitly carves out a niche against named alternatives — 'If you need pretrained models and diverse feature extractors, use deepchem; if you need benchmark datasets, use pytdc' — giving a clear, low-conflict trigger profile.	3 / 3
	Total	11 / 12 Passed

Validation

87%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 2 missing	Warning

	Total	14 / 16 Passed

Repository: aipoch/medical-research-skills
Commit: 7cc5680

Reviewed: 3 days ago

Table of Contents

Discovery Implementation Validation

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.