pyhealth

Comprehensive healthcare AI toolkit for developing, testing, and deploying machine learning models with clinical data. This skill should be used when working with electronic health records (EHR), clinical prediction tasks (mortality, readmission, drug recommendation), medical coding systems (ICD, NDC, ATC), physiological signals (EEG, ECG), healthcare datasets (MIMIC-III/IV, eICU, OMOP), or implementing deep learning models for healthcare applications (RETAIN, SafeDrug, Transformer, GNN).

Quality

81%

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

71%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 body is highly actionable and well-structured for progressive disclosure, with real executable pipelines and properly signaled one-level-deep references. Its main weaknesses are conciseness (duplicated examples and use-case patterns) and missing validation/feedback checkpoints in the training workflow.

Suggestions

Consolidate the 'Quick Start Workflow' and 'Example: Complete Workflow' into a single example to remove ~90 lines of near-duplicate code, and trim the six use cases to 2–3 representative ones that each point to references without restating the pipeline.

Add explicit validation/feedback steps to the training workflow, e.g. reloading the best checkpoint after training ('trainer.load(best_checkpoint)') and a note on handling diverging loss or validation degradation before proceeding to evaluation.

Drop the 'Resources' section that re-lists reference files with word counts, since the 'Detailed Documentation' section already signals each reference with Read-when triggers and Key Topics.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient and assumes Claude's competence, but it carries notable redundancy—both a 'Quick Start Workflow' and a near-duplicate ~90-line 'Complete Workflow' example, six use cases that mostly restate the same reference-reading pattern, and a 'Resources' section that repeats reference word counts—so it could be tightened well below the current length.	3 / 5
Actionability	It provides two fully executable, copy-paste-ready pipelines with real imports and function names (MIMIC4Dataset, split_by_patient, Trainer.train with monitor='pr_auc_score', RETAIN with attention outputs), plus concrete best-practice snippets covering the common cases.	5 / 5
Workflow Clarity	The 5-stage pipeline and numbered use-case steps give a clear sequence, but training/model-saving are batch operations with no explicit validation checkpoints or error-recovery feedback loops (e.g., reloading the best checkpoint, what to do on diverging loss), which caps this dimension at 3 per the rubric.	3 / 5
Progressive Disclosure	The skill offers a clear overview with six well-signaled, one-level-deep reference files—each with a 'Read when' trigger and 'Key Topics' list—and all referenced paths (datasets.md, medical_coding.md, tasks.md, models.md, preprocessing.md, training_evaluation.md) resolve to real files, giving easy navigation.	5 / 5
	Total	16 / 20 Passed

Description

92%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 strong across the board: it concretely lists capabilities, provides explicit use-when triggers, and carves out a distinctive healthcare-AI niche. The only minor gap is trigger-term coverage resting on acronyms rather than broader synonyms.

Dimension	Reasoning	Score
Specificity	The description enumerates many concrete actions and capabilities—'developing, testing, and deploying machine learning models with clinical data'—alongside specific named tasks (mortality, readmission, drug recommendation), coding systems (ICD, NDC, ATC), signal types (EEG, ECG), datasets (MIMIC-III/IV, eICU, OMOP), and model families (RETAIN, SafeDrug, Transformer, GNN), giving comprehensive coverage.	5 / 5
Completeness	It clearly states what the skill does ('Comprehensive healthcare AI toolkit for developing, testing, and deploying machine learning models with clinical data') and explicitly when to use it ('This skill should be used when working with...'), answering both with concrete trigger phrases.	5 / 5
Trigger Term Quality	Natural trigger terms are strong and varied—'electronic health records (EHR)', 'mortality', 'readmission', 'ICD', 'EEG, ECG', 'MIMIC-III/IV'—but coverage leans on acronyms and dataset names and omits a few common synonyms/extensions a user might say, so it sits just below a 5.	4 / 5
Distinctiveness Conflict Risk	The healthcare-AI niche is highly specialized with distinctive triggers (EHR, MIMIC, ICD/NDC/ATC coding, RETAIN/SafeDrug), making it clearly distinguishable from other skills with minimal conflict risk.	5 / 5
	Total	19 / 20 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
metadata_version	'metadata.version' is missing	Warning
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	14 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/biology/pyhealth/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.