Content
38%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 skill body is well-organized by topic and contains genuinely useful concrete artifacts (LaTeX first-page example, schematic command, document structure specs), but it is heavily padded with repeated mandates and duplicated scope sections, lacks a sequenced end-to-end workflow with validation checkpoints, and advertises 14+ bundle files that are missing from the bundle entirely. The dominant failure is verbosity plus broken file references rather than missing domain knowledge.
Suggestions
Actually ship the referenced files (references/, scripts/, assets/) or delete the "See the references/scripts/assets directory" sections — 14 named, non-existent files make the progressive-disclosure structure a dead end.
Merge the three separate statements of the page-1 executive-summary mandate (Document Structure, Output Format, Best Practices) into one section, and collapse "When to Use"/"Integration"/"Key Differentiators" into a single boundary section — this alone would cut hundreds of tokens.
Add an explicit sequenced workflow (gather inputs → run analysis scripts → assemble from template → run validate_cds_document.py → fix and re-validate) with the validation script embedded as a checkpoint.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~530-line body is noticeably verbose with several padded, redundant sections: the page-1 executive summary mandate is stated three times ("CRITICAL REQUIREMENT: All clinical decision support documents MUST begin with a complete executive summary on page 1" in Document Structure, "MANDATORY FIRST PAGE REQUIREMENT" in Output Format, and "First Page Executive Summary (MANDATORY)" in Best Practices); the treatment-plans boundary is repeated across "When to Use" ("Do NOT use this skill for: Individual patient treatment plans"), "Integration", and "Key Differentiators from Treatment-Plans Skill"; and the marketing-style use-case catalogs (Drug Development, Medical Affairs, Clinical Guidelines, Real-World Evidence) restate capabilities already listed. This matches anchor 2 ('noticeably verbose; several unnecessary explanations or padded sections'). Not 1 because it does not explain concepts Claude already knows (no basic PDF/LaTeX tutorials), and not 3 because the duplication is substantial, not just occasional over-explanation. | 2 / 5 |
Actionability | Concrete elements exist: a copy-paste-ready LaTeX first-page example with tcolorbox code, the command "python scripts/generate_schematic.py \"your diagram description\" -o figures/output.png", and named templates. However, the bulk of the body is descriptive capability enumeration ("Evidence-based treatment guidelines for specific disease states", "Subgroup Analyses: Forest plots showing treatment effects") rather than executable instruction, and the referenced scripts and templates (e.g., generate_survival_analysis.py, cohort_analysis_template.tex) do not exist in the bundle, so a user following them hits dead ends. This lands between anchor 2 and anchor 4, fitting anchor 3 ('some concrete guidance but incomplete; missing key details'). | 3 / 5 |
Workflow Clarity | Sub-workflows are sequenced (the schematic steps "1. Generate at minimum ONE schematic... 2. For cohort analyses: include patient flow diagram" and the page-1 → TOC → detailed-sections layout), but there is no overall end-to-end sequence for producing a document (inputs → analysis → assembly → verification). Validation is named as a tool ("validate_cds_document.py - Quality and compliance checks") but never embedded as an explicit checkpoint in a workflow with a fix-and-retry loop, matching anchor 3 ('steps listed but validation gaps; checkpoints missing or implicit'). Not 4 because no coherent document-generation sequence with verification steps exists; not 2 because the sub-processes that are present are clearly defined. | 3 / 5 |
Progressive Disclosure | The body points to "references/", "assets/" (cohort_analysis_template.tex, treatment_recommendation_template.tex, etc.) and "scripts/" (generate_survival_analysis.py, create_waterfall_plot.py, etc.), but none of these files or directories exist in the actual bundle — every reference dangles. Meanwhile all detail (capability catalogs, best practices, five usage examples, GRADE tables) is inlined in a single ~530-line SKILL.md that clearly belongs in separate files. Per the guideline to score against the actual bundle structure, this matches anchor 2 ('content that clearly belongs in separate files is inlined'). Not 1 because section headers provide real navigation and references are one level (not nested); not 3 because the referenced files are absent, so the split promised by the pointers does not actually exist. | 2 / 5 |
Total | 10 / 20 Passed |