Content
82%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.
A strong, highly actionable body: concrete executable commands throughout, a clear numbered core workflow, and good section organization without any bundle overhead. Weaknesses are limited to a filler "When to Use" section, optional rather than enforced validation checkpoints, and a description-duplicating opening line.
Suggestions
Delete the boilerplate "When to Use" section ("Use this skill when tackling tasks related to its primary domain or functionality as described above") and the duplicating opening line — both are removable noise.
Make the availability check an explicit checkpoint with a failure branch: if /is-valid fails, explain what to try next (gated dataset auth via HF_TOKEN, checking /splits for the correct config/split) before proceeding.
Consider moving the full endpoint reference and the upload flows into a references/ file, keeping SKILL.md as the workflow overview plus the most common examples.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean — compact endpoint templates, executable one-liners, and no concept explanations Claude already knows. Minor padding could be trimmed: the closing section "Use this skill when tackling tasks related to its primary domain or functionality as described above" is pure filler, and the opening line duplicates the frontmatter description. | 4 / 5 |
Actionability | Guidance is fully executable: copy-paste-ready curl commands with a real dataset (stanfordnlp/imdb), a working jq pipeline deriving parquet alias paths from cfahlgren1/hub-stats, executable npx commands, and concrete SQL including COPY-based CSV/JSON/Parquet exports. Placeholders like <namespace/repo> are clearly parameter templates, not pseudocode. | 5 / 5 |
Workflow Clarity | The numbered 6-step core workflow gives a clear sequence, pagination continuation logic is specified via response fields (num_rows_total, num_rows_per_page, partial), and the upload flow includes a verify step. It falls short of anchor 5 because validation is only "Optionally validate" with no failure branch (e.g., what to do when /is-valid fails or a dataset is gated). | 4 / 5 |
Progressive Disclosure | No bundle files exist and none are referenced; the single-file body is well-sectioned (Core workflow, Defaults, endpoints, Querying, Uploading) with no dangling references at ~127 lines. Minor gaps: the boilerplate "When to Use" section adds navigation noise, and the full endpoint reference plus upload flows could arguably live in a separate reference file. | 4 / 5 |
Total | 17 / 20 Passed |