cellxgene-census

Query the CELLxGENE Census (61M+ cells) programmatically. Use when you need expression data across tissues, diseases, or cell types from the largest curated single-cell atlas. Best for population-scale queries, reference atlas comparisons. For analyzing your own data use scanpy or scvi-tools.

Quality

72%

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

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/databases/cellxgene-census/SKILL.md

Quality

Content

65%

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

A highly actionable, code-dense body with a clear workflow sequence and well-signaled references. It loses points on conciseness and progressive disclosure because substantial detail (metadata fields, query patterns, best practices) is duplicated inline rather than offloaded to the reference files, and validation checkpoints are not embedded in the batch workflows.

Suggestions

Trim the inline metadata-field list and the per-pattern best-practices that duplicate references/census_schema.md and references/common_patterns.md, replacing them with pointers to those files.

Add inline validation/feedback checkpoints to the batch workflows (e.g., assert cell count before loading, verify batch tensor shape inside the out-of-core loop) instead of relying on a detached Troubleshooting section.

Consolidate the overlapping "When to Use This Skill", "Common Use Cases", and "Key Concepts and Best Practices" sections with the numbered workflow patterns to remove redundancy.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient executable code, but it is padded with redundancy: "When to Use This Skill", "Common Use Cases", and "Key Concepts and Best Practices" repeat scenarios and guidance already embedded in the numbered workflow sections (e.g., is_primary_data filtering and census_version appear both inline and as standalone best practices).	2 / 3
Actionability	It provides fully executable, copy-paste-ready Python across opening, querying, out-of-core iteration, PyTorch training, and scanpy integration, plus concrete install commands and specific value-filter syntax.	3 / 3
Workflow Clarity	A clear numbered sequence (open -> explore -> query -> out-of-core -> ML -> scanpy -> multi-dataset) and a pre-flight size check exist, but validation/feedback loops are detached in a separate Troubleshooting section rather than inline within the batch and out-of-core workflows, so checkpoints are implicit.	2 / 3
Progressive Disclosure	The two reference files are real, one level deep, and well signaled with "When to read" triggers, but the body reproduces metadata-field lists and query patterns that already live in references/census_schema.md and references/common_patterns.md, so content that should be separate remains inline.	2 / 3
	Total	9 / 12 Passed

Description

80%

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 with explicit what/when triggers, natural domain keywords, and excellent distinctiveness via contrast with scanpy/scvi-tools. Its main weakness is voice: the second-person "you need" trigger incurs the rubric's specificity penalty.

Suggestions

Rewrite the trigger clause in third person (e.g., "Use when expression data is needed across tissues, diseases, or cell types") to avoid the second-person specificity penalty.

List multiple distinct concrete actions (e.g., query expression matrices, explore cell/gene metadata, train models via PyTorch, process out-of-core) rather than restating "query" to lift specificity to the top anchor.

Dimension	Reasoning	Score
Specificity	The description names the domain and one concrete action ("Query the CELLxGENE Census (61M+ cells) programmatically"), but not multiple distinct actions; the rubric's mandatory -1 voice penalty applies because the trigger clause uses second person ("Use when you need expression data..."), dropping an otherwise-2 to 1.	1 / 3
Completeness	It explicitly answers both what ("Query the CELLxGENE Census (61M+ cells) programmatically") and when via an explicit "Use when..." trigger clause, satisfying the top anchor.	3 / 3
Trigger Term Quality	Natural domain terms a bioinformatics user would actually say are well covered: "expression data", "tissues, diseases, or cell types", and "single-cell atlas".	3 / 3
Distinctiveness Conflict Risk	A clear niche (curated single-cell atlas) is reinforced by explicit disambiguation ("For analyzing your own data use scanpy or scvi-tools"), making conflict with adjacent skills unlikely.	3 / 3
	Total	10 / 12 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 (511 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/databases/cellxgene-census/SKILL.md
Commit: 52845c3

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.