Content
50%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 has genuinely useful concrete material (CLI examples, parameter table, agency-specific page limits, real reference templates), but it is buried in extensive generic boilerplate that roughly doubles the document's length. Structural issues — a stale file reference, inconsistent section names between commands and the parameter table, and duplicated validation sections — further dilute it.
Suggestions
Cut the boilerplate sections (Risk Assessment, Security Checklist, Evaluation Criteria/Test Cases, Lifecycle Status, and the overlapping Output Requirements/Output Contract/Failure Handling/Error Handling/User Checkpoints/Quick Validation policy blocks) down to one concise guardrails section, keeping only the Quick Check command once.
Fix the reference mismatch: 'references/budget_templates.xlsx' should be 'references/budget_templates.md', and link each reference file from the section that uses it (e.g., Best Practices → review_checklist.md, Specific Aims guidance → specific_aims_examples.md).
Replace the abstract 5-step policy Workflow with a concrete operational sequence for the skill's actual tasks (e.g., generate aims template → fill in with user inputs → validate against review_checklist.md → flag gaps), and reconcile the '--section project_summary' example with the documented section options.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is noticeably verbose: 'Quick Check' and 'Audit-Ready Commands' duplicate the same py_compile validation, and ~10 boilerplate policy sections (Risk Assessment, Security Checklist, Evaluation Criteria, Lifecycle Status with a '2026-03-06' date, Output Requirements/Output Contract, Failure Handling/Error Handling, User Checkpoints, Quick Validation) overlap heavily and add little Claude doesn't already know. It is not a 1 because the core Usage, Parameters, and Agency-Specific sections are concrete, but the padding goes well beyond 'some unnecessary explanation'. | 2 / 5 |
Actionability | Provides concrete, executable commands ('python scripts/main.py --section full --agency NIH --type R01 --output proposal.md'), a full parameter table, and a library usage example. Minor gaps keep it from a 5: 'items=[...]' is a placeholder, and 'Audit-Ready Commands' uses '--section project_summary', which is not among the documented options (aims, significance, approach, budget, full). | 4 / 5 |
Workflow Clarity | A 5-step Workflow exists and a validation entry point is present (py_compile Quick Check, plus fallback handling in Error Handling), but the steps are abstract policy language ('Confirm the user objective', 'Return a structured result') rather than a concrete sequence for drafting/reviewing a proposal section. Checkpoints are partially implicit and the actual template-to-output sequence is never laid out, matching anchor 3 rather than 4. | 3 / 5 |
Progressive Disclosure | References exist and are one level deep with descriptions, but the listed 'references/budget_templates.xlsx' does not exist (the actual file is budget_templates.md), reference files are never linked from the sections that need them (e.g., Best Practices does not point to review_checklist.md), and large amounts of generic policy content are inlined in SKILL.md instead of being split out. This fits anchor 3; not 4 because of the broken reference and buried navigation. | 3 / 5 |
Total | 12 / 20 Passed |