Content
78%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 lean, well-structured instruction set with concrete tool routing, sensible figure-selection rules, a caption template, and an output contract. Its main defects are the absence of worked examples (sample caption, example schematic prompt), no explicit feedback loop after the figure inspection step, and two dangling references to _references files that do not exist in the bundle.
Suggestions
Fix or remove the dangling references to `../_references/section-patterns-and-story.md` and `../_references/submission-and-editorial-readiness.md` — neither file exists in the bundle, so following the skill's own guidance fails.
Add one worked example: a filled-in caption using the 4-part template, and an example prompt for `generate_schematic_figure`, to make the guidance copy-paste ready.
Add an explicit feedback loop after the step-5 inspection, e.g. "If the intended comparison is not obvious, adjust labels or panel layout and re-inspect before writing the caption."
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is ~60 lines of lean imperative rules ("One figure should usually do one argumentative job", "Keep figure text sparse", "Any explanation longer than a short phrase belongs in the caption") with zero padding and no explanation of concepts Claude already knows — every token is domain judgment Claude would not reliably produce on its own, matching the 'lean and efficient' anchor. | 5 / 5 |
Actionability | Concrete tool names (`generate_schematic_figure`, `render_structure_views`, `bash`), an explicit Output Contract ("Return figure intent, output path, script path, source refs, and caption draft"), and a 4-part caption template make the guidance mostly executable, but there are no worked examples — no sample caption and no example schematic prompt — so it falls short of the 'specific examples cover the common cases' anchor. Not 3 because the guidance names real tools and concrete steps rather than pseudocode. | 4 / 5 |
Workflow Clarity | The six-step workflow is clearly sequenced with a gating condition in step 1 ("if the claim is unclear, the figure is not ready") and an inspection checkpoint in step 5, but there is no explicit feedback loop telling Claude what to do when the step-5 inspection fails — 'clear sequence with most checkpoints present; minor validation gaps', not 5. No destructive/batch cap applies. | 4 / 5 |
Progressive Disclosure | Sections are well organized and the two references are one level deep and clearly signaled with purpose ("for figure logic and evidence-first ordering", "before finalizing legends for a journal-facing draft"), but neither referenced file (`../_references/section-patterns-and-story.md`, `../_references/submission-and-editorial-readiness.md`) exists in the bundle — no references/, scripts/, assets/, or _references/ directories are present — so the disclosure promise is unfulfilled, which is more than the 'minor organization gaps' allowed at anchor 4. | 3 / 5 |
Total | 16 / 20 Passed |