Content
45%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 core Procedure is a well-sequenced, reasonably concrete instruction workflow, but it is buried in a large volume of template boilerplate and depends on a references/prompts.md file that is absent from the bundle. The skill's two structural weaknesses — pervasive padding and a broken reference — pull conciseness and progressive disclosure well down.
Suggestions
Ship the referenced `references/prompts.md` file (the body cites it twice for detailed per-step prompts) or remove all references to it — currently the skill instructs Claude to read a file that does not exist.
Delete the boilerplate sections (When to Use, Key Features, Dependencies, Example Usage, Implementation Details, Recommended Workflow, Quick Validation) that restate generic process guidance, keeping only the Procedure, Usage, Output Contract, and validation rules to cut the body to a fraction of its size.
Merge the duplicate 'Procedure' and 'Recommended Workflow' sections into one authoritative sequence, and replace generic checkpoint language ('Run a final validation pass') with concrete pass/fail criteria for the assembled proposal.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly two-thirds of the body is generic boilerplate: 'Use this skill when the request matches its documented task boundary' (circular), a 'Key Features' section that restates the frontmatter description verbatim, an 'Example Usage' block of meta noise ('No packaged executable script was detected'), and a 'Quick Validation' section that says 'No local script validation step is required for this skill'. It is not a 1 because the Procedure section itself is lean and skill-specific; it is below anchor 3 because the padding is pervasive across many sections, not occasional. | 2 / 5 |
Actionability | The Procedure gives genuinely concrete instruction-only guidance — 'Include Model, Groups, Sample, Method, Index, Expected Results', 'Animal experiments must start with modeling', 'a 3-year plan (RMB)' — but the detailed per-step prompts are delegated to `references/prompts.md` ('Refer to [prompts.md](references/prompts.md) for detailed prompts for each step'), and that file does not exist in the bundle. That missing key detail fits anchor 3 rather than anchor 4's 'minor gaps'; it stays above anchor 2 because the in-body steps are specific and executable. | 3 / 5 |
Workflow Clarity | The Procedure is a clearly sequenced 5-step workflow with an explicit final assembly order (Overview → Research Plan → Budget → Yearly Plan) and stated checkpoints ('Ensure consistency between the hypothesis and the designed experiments', 'Run a final validation pass for completeness, consistency, and safety'). Not a 5 because the checkpoints are generic rather than concrete pass/fail criteria, and a second 'Recommended Workflow' section duplicates the Procedure creating mild ambiguity about the authoritative path; well above anchor 3 since validation steps are present and explicit. | 4 / 5 |
Progressive Disclosure | The body claims 'Reference material available in `references/`' and twice directs the reader to `references/prompts.md`, but no references/ directory or prompts.md exists anywhere in the bundle — the skill's only detailed reference is a dangling link, which breaks navigation entirely when followed. Additionally, large amounts of generic boilerplate are inlined in SKILL.md that belong in separate files or nowhere. It escapes anchor 1 only because sections are clearly labeled and the (broken) reference design is one level deep and clearly signaled. | 2 / 5 |
Total | 11 / 20 Passed |