deepchem

Molecular ML with diverse featurizers and pre-built datasets. Use for property prediction (ADMET, toxicity) with traditional ML or GNNs when you want extensive featurization options and MoleculeNet benchmarks. Best for quick experiments with pre-trained models, diverse molecular representations. For graph-first PyTorch workflows use torchdrug; for benchmark datasets use pytdc.

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

78%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.

A thorough, highly actionable reference with executable examples, decision aids, and well-signaled one-level-deep bundle files. Its main weakness is moderate verbosity: background explanation and duplicate installation content could be trimmed to respect the context budget.

Suggestions

Cut or shorten the Overview paragraph and 'When to Use This Skill' bullets that restate general library/domain knowledge Claude already has.

Consolidate the duplicate install instructions (Issue 4 'Import Errors' and 'Installation Notes') into a single block to save tokens.

Add an explicit evaluation feedback loop (e.g. 'if ROC-AUC < baseline, increase epochs / switch architecture / check scaffold split') to the model-training workflows.

Dimension	Reasoning	Score
Conciseness	Mostly efficient with concrete code, tables, and decision trees, but the Overview and 'When to Use This Skill' sections restate library knowledge Claude already has, and installation instructions appear twice (Issue 4 and Installation Notes).	3 / 5
Actionability	Abundant copy-paste-ready code, full CLI invocations with flags, featurizer/model selection tables, and a decision tree that cover the common molecular-ML cases end to end.	5 / 5
Workflow Clarity	Workflows A/B/C use numbered steps terminating in an evaluation checkpoint, plus pitfall-driven recovery advice; however there is no explicit validate->fix->retry feedback loop around training/evaluation.	4 / 5
Progressive Disclosure	Body keeps overview and key examples inline while clearly signaling one-level-deep references (references/api_reference.md, references/workflows.md) and three real scripts in scripts/, all of which exist as bundle files.	5 / 5
	Total	17 / 20 Passed

Description

87%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.

A strong description that crisply states the skill's purpose, gives explicit 'Use for'/'Best for' triggers, and includes routing guidance to avoid adjacent skills. Minor tightening of padded phrasing and a few more natural synonyms would push it to fully clean.

Suggestions

Drop vague qualifiers like 'diverse featurizers' and 'diverse molecular representations' in favor of naming one or two specific representations (e.g. ECFP fingerprints, molecular graphs).

Add high-frequency natural terms users say ('SMILES', 'solubility', 'drug discovery') to broaden trigger coverage.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions (property prediction, ADMET, toxicity, traditional ML, GNNs, featurization, MoleculeNet benchmarks, pre-trained models) but pads with vague phrasing like 'diverse featurizers' and 'diverse molecular representations', leaving it just shy of cleanly comprehensive.	4 / 5
Completeness	Explicitly answers both what ('Molecular ML with diverse featurizers and pre-built datasets') and when ('Use for property prediction (ADMET, toxicity)... Best for quick experiments...'), with concrete trigger phrases and even negative routing guidance.	5 / 5
Trigger Term Quality	Strong natural keywords users would say (ADMET, toxicity, property prediction, MoleculeNet, GNNs, featurization, pre-trained models) plus disambiguating names (torchdrug, pytdc); missing a few common variants like 'drug discovery', 'SMILES', or 'solubility'.	4 / 5
Distinctiveness Conflict Risk	Clear DeepChem niche with distinct molecular-ML triggers and explicit disambiguation (torchdrug for graph-first PyTorch, pytdc for benchmark datasets), minimizing conflict risk.	5 / 5
	Total	18 / 20 Passed

Validation

81%

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 — 13 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (601 lines); consider splitting into references/ and linking	Warning
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	13 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/chemistry/deepchem/SKILL.md
Commit: 3a6c3a9

Reviewed: 1 day 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.