CtrlK
BlogDocsLog inGet started
Tessl Logo

torchdrug

Graph-based drug discovery toolkit. Molecular property prediction (ADMET), protein modeling, knowledge graph reasoning, molecular generation, retrosynthesis, GNNs (GIN, GAT, SchNet), 40+ datasets, for PyTorch-based ML on molecules, proteins, and biomedical graphs.

56

Quality

62%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./bundled/skills/torchdrug/SKILL.md
SKILL.md
Quality
Evals
Security

Quality

Content

62%Weight 40%Scale 1-5

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

The skill body is well-structured with strong progressive disclosure to real reference files and mostly executable code, but it is somewhat verbose due to repeated capability lists and lacks explicit validation checkpoints in its multi-step workflows, capping workflow clarity.

Suggestions

Trim the redundant Overview and Summary capability lists since the Core Capabilities and Summary sections already enumerate the same domains and reference files.

Add explicit validation/evaluation checkpoints to the workflows (e.g., 'After training: evaluate on test split and confirm AUROC > baseline before declaring success') to lift workflow clarity above 3.

Remove restated explanations of basic concepts (e.g., what scaffold splitting or AUROC is) where Claude can already infer them from the code.

DimensionReasoningScore

Conciseness

Mostly efficient with concrete code and bullets, but the body repeats capability lists already covered by reference files and includes some restated overview prose (e.g., the Overview and Summary restate the same domain list), adding tokens Claude could derive.

3 / 5

Actionability

Provides multiple copy-paste-ready code blocks (BBBP/GIN training loop, RDKit conversion, AlphaFold/ESM loading, Lightning wrapper) with real imports and concrete calls, with only minor gaps such as unspecified hyperparameters and missing evaluation code.

4 / 5

Workflow Clarity

Workflows are clearly numbered and sequenced with navigation pointers, but they list steps as plain descriptions without explicit validation checkpoints or feedback loops for batch/training operations, which the rubric caps at 3 for missing validation.

3 / 5

Progressive Disclosure

Excellent one-level-deep structure: the body is a concise overview with clearly signaled references to eight real files in references/, each named in context with what it contains, and navigation cheat sheets and a summary make discovery easy.

5 / 5

Total

15

/

20

Passed

Description

62%Weight 40%Scale 1-5

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

The description is specific and distinctive with strong domain keyword coverage, but it omits any explicit 'when to use' trigger guidance, which caps completeness at 3 and slightly limits trigger_term_quality. Adding a 'Use when...' clause naming natural user phrases would lift the two weakest dimensions.

Suggestions

Add an explicit trigger clause, e.g. 'Use when working with molecules, SMILES strings, protein structures, retrosynthesis, or graph-based drug discovery in PyTorch.'

Include natural user-side terms and synonyms such as 'SMILES', 'molecules', 'drug-target binding', and 'chemical reactions' alongside the technical vocabulary.

Lead with concrete action verbs ('Predict molecular properties', 'Model proteins', 'Plan retrosynthesis') rather than a noun-tag list to strengthen specificity.

DimensionReasoningScore

Specificity

Lists several concrete capabilities (molecular property prediction/ADMET, protein modeling, knowledge graph reasoning, molecular generation, retrosynthesis) plus specific model families (GIN, GAT, SchNet) and a dataset count, but phrase structure reads as a tag list rather than discrete action verbs.

4 / 5

Completeness

The 'what' is clearly and concretely answered, but there is no 'Use when...' or equivalent explicit trigger clause describing when Claude should invoke this skill, which caps completeness at 3 per the rubric guideline.

3 / 5

Trigger Term Quality

Strong domain keywords (drug discovery, ADMET, retrosynthesis, GNNs, PyTorch) plus concrete model names, but lacks the natural user-side phrasing and synonyms (e.g., 'SMILES', 'molecules', 'drug-target') that users would actually say.

4 / 5

Distinctiveness Conflict Risk

The TorchDrug niche (graph-based drug discovery with named GNN architectures and biomedical tasks) is highly specific with clear, distinctive triggers and minimal overlap risk with other skills.

5 / 5

Total

16

/

20

Passed

Validation

100%

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

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
foryourhealth111-pixel/Vibe-Skills
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.