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

81%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

72%

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

A highly actionable, well-structured skill with good progressive disclosure to real reference files and scripts. Weak spots are mild verbosity from duplicated examples across sections and workflows that lack explicit validation feedback loops.

Suggestions

Trim redundant code between "Core Capabilities", "Typical Workflows", and "Example Scripts" so each example earns its place.

Add an explicit validation/retry checkpoint inside the workflows (e.g. after training: evaluate on validation, if metric below threshold adjust epochs/dropout and retrain) rather than relegating all recovery to "Common Pitfalls".

Dimension	Reasoning	Score
Conciseness	The body is mostly actionable with little concept-explanation fluff, but it is long (~600 lines) and the "Typical Workflows" and "Example Scripts" sections re-demonstrate code already shown in "Core Capabilities", so it could be tightened.	2 / 3
Actionability	Packed with concrete, executable code (loaders, featurizers, models, evaluation), a featurizer decision tree, a model-selection table, and copy-paste-ready CLI invocations for the bundled scripts.	3 / 3
Workflow Clarity	Workflows A/B/C are clearly numbered and end with an evaluate step, but they lack explicit in-workflow validation checkpoints or validate-fix-retry feedback loops; recovery guidance lives only in the separate "Common Pitfalls" section.	2 / 3
Progressive Disclosure	SKILL.md is an overview that clearly signals one-level-deep, real bundle files ("See references/api_reference.md", "See references/workflows.md") plus the scripts/ directory, all of which exist on disk; content is appropriately split.	3 / 3
	Total	10 / 12 Passed

Description

90%

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, specific description with explicit trigger guidance and good distinctiveness from neighboring skills. The only blemish is second-person voice ("when you want"), which costs it one specificity point under the rubric.

Suggestions

Rewrite in third person to avoid the voice penalty, e.g. "Use for property prediction (ADMET, toxicity) with traditional ML or GNNs when extensive featurization options and MoleculeNet benchmarks are needed."

Dimension	Reasoning	Score
Specificity	Lists many concrete actions ("property prediction (ADMET, toxicity)", "traditional ML or GNNs", "MoleculeNet benchmarks", "pre-trained models"), which is baseline-3 specificity, but the second-person phrasing "when you want extensive featurization options" triggers the rubric's -1 voice penalty.	2 / 3
Completeness	Explicitly answers both what ("Molecular ML with diverse featurizers and pre-built datasets") and when ("Use for property prediction... when you want extensive featurization options and MoleculeNet benchmarks"), an explicit Use-when trigger clause.	3 / 3
Trigger Term Quality	Good coverage of natural domain terms a user would say — "property prediction", "ADMET", "toxicity", "GNNs", "pre-trained models", "MoleculeNet benchmarks" — with common variations, not just jargon.	3 / 3
Distinctiveness Conflict Risk	Clear niche (molecular ML via DeepChem) and it explicitly disambiguates from adjacent skills ("For graph-first PyTorch workflows use torchdrug; for benchmark datasets use pytdc"), so it is unlikely to trigger for the wrong skill.	3 / 3
	Total	11 / 12 Passed

Validation

93%

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 — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (603 lines); consider splitting into references/ and linking	Warning

	Total	15 / 16 Passed

Repository: K-Dense-AI/scientific-agent-skills
Path: skills/deepchem/SKILL.md
Commit: 757b63b

Reviewed: about 11 hours 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.