Content
75%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 is an efficient, well-structured conceptual overview that delegates detail to a real, well-organized reference bundle — exactly the right shape for a skill about a library outside training data. It falls short of top marks only in small ways: no inline executable snippet, an unnamed validation command, and two bundle files left undiscoverable from the body.
Suggestions
Add a one-line pointer to references/faq.md (and README.md) in the body — e.g. under the pre-code reading list — so every bundle file is discoverable from SKILL.md.
Include the concrete command or fixture-invocation used by the parse→serialize round-trip gate so the validation checkpoint is executable, not just named.
Add a minimal copy-paste skeleton (e.g. one Plot.define plus its extension bundle with command, key binding, and MarkdownSpec) or name which file in references/examples/ covers the most common task first, so the body's guidance is fully executable before jumping to references.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and fluff-free and assumes Claude's competence (it never explains what an editor or schema is), but it carries time-sensitive statements ("released in 2026-07", "Wordgard is 0.x") outside any old-patterns/deprecated section — the 2026-07 date is load-bearing for "never guess its API", yet still a trim candidate per the time-sensitivity guideline, keeping it just below the every-token-earns-its-place anchor. | 4 / 5 |
Actionability | Concrete API names appear throughout ("Plot.define", "wg.dispatch(spec)", "Command.handler", "doc.resolve(pos)", "scheduleDOMRead/scheduleDOMWrite") alongside exact instructions to grep the installed typings and read specific example files, but the body itself contains no copy-paste-ready code block, so it lands on "mostly executable guidance, minor gaps" rather than fully executable. | 4 / 5 |
Workflow Clarity | "Before writing any Wordgard code" gives a clear numbered 1–4 reading sequence, and validation is addressed via the "parse→serialize round-trip gate" and golden-corpus fixtures — but the actual command or script for running that validation is never given, leaving a checkpoint named but not executable. | 4 / 5 |
Progressive Disclosure | The body is a genuine overview with well-signaled, one-level-deep references that all resolve to real files (guide.md, migrating-from-prosemirror.md, examples/*.md, ATTRIBUTION.md, CHANGELOG.md); however references/faq.md and references/README.md exist in the bundle yet are never pointed to, a small navigation gap that keeps it below the top anchor. | 4 / 5 |
Total | 16 / 20 Passed |