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 the writing process and the LaTeX report package, with a clearly sequenced manuscript workflow. Its weaknesses are heavy padding with general-knowledge material and a progressive-disclosure structure whose referenced files are absent from the bundle while their would-be content is inlined.
Suggestions
Move Section 10 (field-specific terminology), Section 6 (writing principles), and the reporting-guideline lists into the referenced files and reduce each to a one-line pointer, cutting hundreds of tokens of knowledge Claude already has.
Ship the referenced bundle files (references/*.md, scripts/generate_schematic.py, scripts/generate_image.py, assets/scientific_report.sty and templates) or remove the dangling references so pointers resolve.
Replace the rigid figure-count minimum table with a short heuristic, and trim the repeated MANDATORY/EXTENSIVELY figure-generation exhortations to a single concise instruction.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~720-line body spends large stretches on knowledge Claude already has — ~90 lines of field-specific terminology conventions (italic gene symbols, binomial nomenclature, SI units, person-first language), generic clarity/conciseness/accuracy/objectivity principles, common figure types, and arbitrary figure-count tables ('Market Research: 20 minimum, 25-30 recommended'). Genuinely skill-specific material (the two-stage writing process, scientific_report.sty commands) keeps it above anchor 1, but the padding is substantial. | 2 / 5 |
Actionability | Guidance is mostly executable: copy-paste bash commands for figure generation, complete LaTeX blocks for box environments, tables, and the document skeleton, the xelatex compile command, and a fully worked outline-to-prose conversion example. Minor gaps come from deferring to external skills (research-lookup, venue-templates) without local detail. | 4 / 5 |
Workflow Clarity | The 'Workflow for Manuscript Development' section sequences four numbered stages, the two-stage outline-to-prose process is unambiguous, and Stage 3 revision serves as a validation checklist (verify citations, check word counts, confirm guideline adherence). It falls short of 5 because checkpoints lack explicit feedback loops for re-checking after fixes. | 4 / 5 |
Progressive Disclosure | References are well-signaled and one level deep (References/Assets sections at the end), but the actual bundle contains none of them — references/, scripts/, and assets/ directories do not exist, so all 11 referenced paths (6 references/*.md, 3 assets/*, 2 scripts/*.py) are dangling. Additionally, ~200+ lines of inline material (field-specific terminology, writing principles, guideline lists) clearly belongs in those separate reference files. | 3 / 5 |
Total | 13 / 20 Passed |