scgpt

Use scGPT-style single-cell foundation model workflows. Use when a task asks for single-cell embeddings, perturbation prediction, cell annotation, batch transfer, or gene-program analysis.

Quality

89%

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

86%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 tight, well-structured workflow-oriented skill that assumes Claude's competence and stays within budget. The main weakness is actionability: step 2's verification target is unspecified and the entirely instruction-based guidance lacks concrete package/checkpoint names or executable examples.

Suggestions

In step 2, name the concrete artifacts to verify (e.g. the scGPT checkpoint repo/URL and the required pip/conda packages) so the verification is executable rather than abstract.

Add a brief error-recovery feedback loop, e.g. 'If a checkpoint is unavailable or a comparison mismatches, re-check preprocessing/identifiers and retry before reporting.'

Optionally include a minimal copy-paste snippet for loading an scGPT checkpoint and producing embeddings, to lift actionability from mostly-executable to fully-executable.

Dimension	Reasoning	Score
Conciseness	The body is lean and token-efficient — a one-line intro, five terse workflow steps, and a closing guardrail — with no padding and no explanation of concepts Claude already knows.	5 / 5
Actionability	Steps give concrete checklists of what to record, save, and compare against (e.g. 'source metadata, marker genes, perturbation controls, and literature'), but step 2 leaves 'model/checkpoint/package availability' unspecified and no executable code or package names are provided.	4 / 5
Workflow Clarity	A clearly numbered five-step sequence includes checkpoints (verify-before-running in step 2, compare-against-ground-truth in step 4, and a closing guardrail), but there is no explicit error-recovery feedback loop (validate -> fix -> retry).	4 / 5
Progressive Disclosure	At ~13 lines with no need for external references, the body is well-organized into a title, intro, workflow list, and guardrail, satisfying the simple-skill exception for progressive disclosure.	5 / 5
	Total	18 / 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.

A strong, concise description that clearly defines a niche domain and provides explicit, concrete trigger guidance for when to invoke the skill. The only minor gap is the absence of a few common synonyms like 'scRNA-seq' and 'batch integration'.

Dimension	Reasoning	Score
Specificity	The 'when' clause lists multiple concrete analysis actions — 'single-cell embeddings, perturbation prediction, cell annotation, batch transfer, or gene-program analysis' — giving comprehensive coverage of the scGPT-style workflow domain.	5 / 5
Completeness	It explicitly states both the 'what' ('Use scGPT-style single-cell foundation model workflows') and the 'when' ('Use when a task asks for...') with concrete trigger phrases.	5 / 5
Trigger Term Quality	Natural domain keywords (embeddings, perturbation prediction, cell annotation, batch transfer, gene-program) are present, but common synonyms such as 'scRNA-seq' and 'batch integration' are missing.	4 / 5
Distinctiveness Conflict Risk	The scGPT framing plus single-cell-specific triggers carve a clear niche with minimal overlap risk against non-single-cell or non-foundation-model skills.	5 / 5
	Total	19 / 20 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: companion-inc/feynman
Path: skills/scgpt/SKILL.md
Commit: 95c51d9

Reviewed: about 9 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.