Content
67%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a thorough, actionable guide to querying the CELLxGENE Census with well-signaled reference files, but it is held back by noticeable redundancy: best-practice guidance and use-case code repeat material already covered in the numbered workflow patterns and in common_patterns.md.
Suggestions
Collapse the 'Common Use Cases' section or convert it to cross-references, since its four code examples largely duplicate the 'Core Workflow Patterns' above.
Trim the 'Key Concepts and Best Practices' section to only items not already covered inline (e.g., remove the repeated 'Always Filter for Primary Data', 'Specify Census Version', 'Select Only Needed Columns', and 'Use tissue_general' subsections), pointing to common_patterns.md instead.
Either define or clearly mark as illustrative the placeholder symbols in the ML examples (model, criterion, optimizer, num_epochs, process_batch) so the code is unambiguously executable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The content is largely API-specific (get_anndata, obs_value_filter, is_primary_data, tissue_general) rather than explaining basics Claude knows, but the 'Key Concepts and Best Practices' and 'Common Use Cases' sections duplicate code and guidance already shown in the numbered 'Core Workflow Patterns' and present in common_patterns.md, so it could be tightened considerably. | 3 / 5 |
Actionability | It provides concrete, mostly copy-paste-ready code for the core query APIs (get_anndata, get_obs, get_var, axis_query, experiment_dataloader) with specific filter syntax; minor gaps remain where ML examples reference undefined placeholders like model, num_epochs, and process_batch(). | 4 / 5 |
Workflow Clarity | The process is clearly sequenced (open census -> explore metadata -> query at small/large scale -> integrate/ML) with an explicit 'Estimate Query Size Before Loading' checkpoint that guards memory, and a troubleshooting section; this is a read/query skill so the destructive-cap does not apply, though post-query validation is only implicit. | 4 / 5 |
Progressive Disclosure | Two real one-level-deep reference files (census_schema.md, common_patterns.md) are signaled in a 'Reference Documentation' section with 'When to read' navigation guidance, but the body inlines substantial patterns and best practices that overlap with common_patterns.md, leaving a minor organization gap. | 4 / 5 |
Total | 15 / 20 Passed |