Content
64%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A solid API reference skill that provides concrete, actionable endpoint patterns and useful curl/npx examples. Its main weaknesses are scope creep (dataset creation/upload in a read-only viewer skill), missing error-handling/validation checkpoints in workflows, and lack of progressive disclosure via supporting files. The filter predicate syntax and search query examples are notably absent.
Suggestions
Add a concrete example for `/filter` showing actual `where` predicate syntax (e.g., `where=column1>5`) and `/search` with a real query string.
Add error-handling guidance: what to do when `/is-valid` returns false, when responses include `partial: true`, or when 401/403 is returned for gated datasets.
Consider removing or splitting the 'Creating and Uploading Datasets' section into a separate file, as it contradicts the skill's stated read-only scope.
Remove the generic 'When to Use' footer—it adds no actionable information.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Generally efficient with minimal over-explanation. The API endpoint listing is compact and useful. Minor verbosity in the 'Creating and Uploading Datasets' section which goes beyond the stated skill scope (read-only Dataset Viewer API), and the 'When to Use' footer is a filler sentence that adds no value. | 4 / 5 |
Actionability | Provides concrete curl commands, npx invocations, and specific endpoint URL patterns that are copy-paste ready. Minor gaps: the `/filter` predicate syntax is not illustrated with an example (just mentions 'predicate syntax in `where`'), and the `/search` endpoint lacks a concrete example query. | 4 / 5 |
Workflow Clarity | The core workflow lists a clear 6-step sequence for exploration, and the pagination pattern is well-illustrated. However, there are no validation checkpoints or error-handling guidance (e.g., what to do if `/is-valid` returns false, how to handle 401/403 for gated datasets, or what to do when `partial` is true). The upload section introduces a destructive-adjacent operation (creating/uploading datasets) without verification beyond a single curl check. | 3 / 5 |
Progressive Disclosure | Content is organized into logical sections with clear headers, but everything is inlined in a single file. The SQL export section, the upload workflows, and the parquet querying details could benefit from being split into separate referenced files. No bundle files exist to offload detail, and no references to external docs are provided. | 3 / 5 |
Total | 14 / 20 Passed |