Content
63%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.
Excellent structure and navigation with genuinely useful agency-specific facts, concrete limits, and a runnable figure command, but the body carries substantial generic padding — an entirely skippable 'Common Mistakes' section, a mostly generic phase-by-phase activity checklist, and a motivational closing note — that a competent model does not need. Trimming that material would make the skill markedly more token-efficient without losing any unique value.
Suggestions
Cut the 'Common Mistakes to Avoid' section (~40 lines) down to grant-specific traps that are not obvious (e.g., archived calls not establishing live funding, modular budget consortium F&A exclusions) or move it to a reference file.
Compress the five-phase workflow to the non-obvious verification steps (institutional submitter routing, receipt/validation confirmation, exported-file page-limit checks) and drop generic activities like 'Proofread all materials' and 'Assemble team'.
Remove the motivational 'Final Note' paragraph and the generic 'Integration with Other Skills' list, or fold the few actionable integration pointers (e.g., scientific-schematics with --doc-type grant) into the Visual Enhancement section where the command already appears.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Several sections restate knowledge Claude already has or are pure padding: the ~40-line 'Common Mistakes to Avoid' section lists generic failure modes like 'Excessive Jargon', 'Verbosity', 'Wrong Font or Margins', and 'Late Submission'; the five-phase workflow is a generic activity checklist ('Proofread all materials', 'Assemble team and confirm collaborations'); and the closing 'Final Note' ('Grant writing is both an art and a science... persistence and revision are key to funding success') is motivational filler. This matches anchor 2 ('noticeably verbose; several unnecessary explanations or padded sections') — there is real agency-specific content inline (page limits, modular budget figures, resubmission policy, PAPPG supplements), so it is above anchor 1's pervasive padding, but the padding is more than the 'some unnecessary explanation' of anchor 3. | 2 / 5 |
Actionability | Quotes: '15-page project description limit (most programs; includes Results from Prior NSF Support, max 5 pages)', 'Specific Aims up to 1 page and Research Strategy up to 12', 'modular budgets use $25,000 increments, up to $250,000 in direct costs per budget period', a runnable command 'python skills/scientific-schematics/scripts/generate_schematic.py "project timeline with Year 1-3 milestones" -o figures/timeline.png --doc-type grant', and the concrete procedure 'Build a requirement -> source section -> draft location -> gap matrix before judging compliance'. For an instruction-only skill this is mostly executable guidance with specific numbers, URLs, and commands, matching anchor 4. It is below anchor 5 because the multi-phase workflow and some sections give topic checklists ('Develop preliminary data (if needed)', 'Write broader impacts') without the concrete steps, which live in references. | 4 / 5 |
Workflow Clarity | The sequence is clear and checkpointed: establish the governing opportunity first ('record agency, opportunity identifier and amendment... due date/time zone... review criteria and source URL/retrieval date'), build the requirement->gap matrix, then five sequenced phases with activities and outputs each, ending in 'Confirm agency receipt, validation status and the assembled application'. Validation checkpoints are present ('Inspect the final exported files; Markdown text does not establish page compliance'; 'Submit 24-48 hours before deadline'), and there is an error-recovery branch ('If the official package is unavailable, provide a bounded drafting review and list unresolved rules'). This matches anchor 4 ('clear sequence with most checkpoints present; minor validation gaps'); it misses anchor 5 because there is no explicit validate->fix->retry loop or checklist for the drafting/compliance phases themselves. The skill is not destructive/batch, so the workflow cap does not apply. | 4 / 5 |
Progressive Disclosure | The body is a clear overview with well-signaled, one-level-deep references, all of which exist in the bundle: 'Section-by-section guidance for every standard proposal component... is in references/core_components.md', a 'Review Criteria, Writing Principles, and Proposal Types' section pointing to review_criteria.md, writing_principles.md, and proposal_types_and_resubmission.md, a 'Resources' section listing each per-agency reference with what it covers, a 'Templates and Assets' section, and even a references/README.md navigation map. Cross-links between references (e.g., review_criteria.md -> nsf_guidelines.md) are lateral pointers among files that each hold their own content, not nested 2+ level chains. This matches anchor 5 ('clear overview with well-signaled one-level-deep references; content appropriately split; easy navigation'). | 5 / 5 |
Total | 15 / 20 Passed |