Content
88%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.
A lean, highly actionable skill body: concrete tool invocations with params and constraints, a validated and authorization-gated publishing workflow, and a useful common-errors list. The only gaps are the abbreviated command lists for material/article management and the inlined API reference in a single file.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every tool entry is a dense signature with params, return shape, and constraints (temp-dir restriction, 2 MB / 10 MB limits); there is no explanation of concepts Claude already knows and no padded sections. Not score 4 because no section can be trimmed without losing non-obvious operational detail. | 5 / 5 |
Actionability | Eleven capabilities give copy-paste-ready `limb_invoke_tool({ nodeId, command, params })` calls covering the common cases, but the 素材管理 and 已发布文章 bullets list bare command names ("list_material、get_material_count、delete_material") without the full invocation format, leaving minor gaps. Not score 5 because those entries are not directly executable as written. | 4 / 5 |
Workflow Clarity | The 发文流程 section gives a clear 7-step sequence with explicit validation (step 6 "list_drafts 回读" checking media_id, title, author, digest, cover, and first-screen subtitle), a pre-flight node check, an authorization gate for the destructive publish step ("仅在用户明确授权正式发表时发布"), and a 常见错误 section for error recovery. | 5 / 5 |
Progressive Disclosure | Well-organized single-file structure with clear section headers (使用前检查, 接口发现, 核心能力, 发文流程, 常见错误, 限制) and no nested or buried references. Not score 5 because the ~35-line tool listing (核心能力) is the kind of API reference that could be split into a separate file now that the skill exceeds 50 lines. | 4 / 5 |
Total | 18 / 20 Passed |