Content
57%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.
An extremely lean overview with clearly signaled references, but the actual operating detail lives in reference files absent from the provided bundle. Standing alone, the body offers principles rather than executable guidance (no tools, commands, or examples) and only an implicit workflow sequence.
Suggestions
Name the concrete deterministic tools and output formats (e.g., matplotlib/pgfplots/TikZ, SVG/PDF) and include one short executable snippet or command for the common data-plot case, replacing the abstract 'Use deterministic editable tools for data and source text'.
Ship the referenced workflow/revision files inside the skill's own references/ directory and use bundle-relative paths (references/scientific-figure-workflow.md) so the links resolve as bundled.
Enumerate the workflow steps (identify → open data/context → design → render → review) and add an explicit fix-and-re-render loop after the final-size review checkpoint.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~200-word body is dense imperative guidance with zero padding — no concept explanations and no restating of what Claude already knows; every sentence directs action ('Open the real data, chosen existing figure and manuscript context when available', 'Review labels, composition and legibility in the final-size render'). Not 4: there is no over-explanation anywhere that could be trimmed. | 5 / 5 |
Actionability | Guidance stays high-level throughout — 'Match layout and marks to the scientific comparison', 'Use deterministic editable tools for data and source text' — with no named tools, commands, formats, or examples; the executable detail is delegated to reference files that are not present in the bundle. Not 3: nothing reaches even pseudocode-level concreteness; not 1: it instructs with specific deliverables ('Deliver the actual vector or editable source, rendered output, and caption with panel meanings') rather than merely describing. | 2 / 5 |
Workflow Clarity | A rough sequence is readable across the paragraphs (identify → open context → design → review render → deliver) and two checkpoints appear ('Review labels, composition and legibility in the final-size render'; attach figures 'for the user's review'), but steps are never enumerated and there is no error-recovery loop after the review. Not 4: the sequence is implicit and checkpoints are partial; not 2: the flow is coherent and validation checkpoints do exist. | 3 / 5 |
Progressive Disclosure | The body is a short, well-signaled overview with two clearly labeled links ('scientific figure workflow', 'figure revision decisions'), but the referenced paths (../../skill/references/scientific-figure-workflow.md, ../../skill/references/figure-revision-decisions.md) point two levels outside the skill's directory and no references/ directory exists in the bundle, so the disclosure chain does not resolve as shipped. Not 4: broken/external reference paths are more than a minor organization gap; not 2: references are clearly signaled, not buried, and no separate-file content is inlined. | 3 / 5 |
Total | 13 / 20 Passed |