Content
56%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 delivers strong, concrete guidance for its core workflow — the two-stage outline-to-prose process with a worked example, plus executable LaTeX for professional report formatting — and points to six real, substantive reference files. Its main weaknesses are verbosity (repeated mandates, a large inline terminology taxonomy of knowledge Claude already has, LaTeX detail duplicating an asset guide) and two broken script references, which together hold the conciseness and progressive-disclosure dimensions down.
Suggestions
Move Section 10 (field-specific terminology, ~90 lines) into a reference file such as 'references/field_terminology.md' and keep a 5-line summary of the principle in SKILL.md.
Trim the LaTeX detail in Section 8 (box environments, table example, commands table) to a short starter snippet plus pointers to assets/REPORT_FORMATTING_GUIDE.md and references/professional_report_formatting.md, which already cover it.
Remove or fix the references to scripts/generate_schematic.py and scripts/generate_image.py (the scripts/ directory does not exist in this bundle), and consolidate the 'never bullet points' mandate to a single stated location instead of repeating it in the Overview, Section 7, and the Workflow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~690-line body has several padded sections: the 'never bullet points' mandate is repeated in the Overview, Section 7, and the Workflow; Section 10 is a ~90-line taxonomy of field terminology conventions (gene italics, SI units, binomial nomenclature) that Claude already knows; and Sections 6/11 restate generic writing principles and common pitfalls. It is above 1 because much of the material (CONSORT/STROBE/PRISMA mappings, venue differences, the .sty workflow) is genuinely domain-specific rather than basic-library or file-format explanation. | 2 / 5 |
Actionability | Guidance is mostly executable: a worked outline-to-prose example, copy-paste LaTeX for box environments, tables, and notation commands, and an XeLaTeX compile command. It falls short of 5 because the schematic generation commands invoke 'scripts/generate_schematic.py' and 'scripts/generate_image.py' which do not exist in the bundle, and several sections defer to external skills ('scientific-schematics', 'generate-image', 'venue-templates') without bundled instructions. | 4 / 5 |
Workflow Clarity | The 'Workflow for Manuscript Development' gives a clear 4-stage numbered sequence, and Stage 3 (Revision) provides verification checkpoints ('Verify all citations are accurate', 'Confirm adherence to reporting guidelines', 'Check word counts'). It is not 5 because checkpoints are advisory prose rather than explicit validate-and-fix feedback loops, and the drafting order (figures first, Methods first) is stated without guidance on recovering when a stage fails. | 4 / 5 |
Progressive Disclosure | References are well signaled — all six 'references/*.md' and three 'assets/*' files exist, are pointed to at point of use, and are summarized in dedicated sections — but content that clearly belongs in reference files is inlined: the ~90-line field-terminology taxonomy (Section 10) and the detailed LaTeX box/table/commands documentation (Section 8) duplicating assets/REPORT_FORMATTING_GUIDE.md. Two referenced script paths do not exist in the bundle, and the venue-templates pointer lists file names Claude cannot navigate from this bundle. | 3 / 5 |
Total | 13 / 20 Passed |