Content
67%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 well-structured, technically dense skill body with concrete executable commands, real bundle files, and an explicit validation gate before downstream consumption. Main weaknesses are the generic common-constraints boilerplate adding tokens and inline schema descriptions instead of clean links to the existing reference files.
Suggestions
Move the verbatim "公共硬约束" boilerplate into a shared reference file and link to it, or trim to only the constraints specific to this skill, to improve conciseness.
Replace the inline candidate-schema field listing with a markdown link to references/paper-schema.md (and bundle-schema.md) so the body points to rather than duplicates the detail.
Make the validate→fix→retry loop explicit in the 校验/失败与恢复 sections (e.g., 'if validation fails, fix and re-run validate before review consumption') to reach the top workflow_clarity anchor.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient and assumes Claude's competence (no basic-concept padding), but the generic "公共硬约束" boilerplate block and some repeated constraint prose add tokens that could be trimmed; not a 4 because the duplicated common-constraints section is noticeable padding, not a 2 because the core technical content is dense and relevant. | 3 / 5 |
Actionability | The 执行步骤 section gives concrete, copy-paste-ready bash commands with real flags (run/enrich-abstracts/validate), but a substantial portion of the body is constraint prose rather than executable guidance; not a 5 because the common-constraints and Skill 专属约束 sections stay abstract. | 4 / 5 |
Workflow Clarity | A clear sequence is present (输入 → 执行步骤 → 输出 → 校验 → 失败与恢复) with an explicit validation checkpoint (the validate subcommand, "review 消费前必须通过"); not a 5 because the validate→fix→retry feedback loop is implied rather than spelled out as an explicit retry loop. | 4 / 5 |
Progressive Disclosure | Structure is clear with well-labeled sections and real one-level-deep bundle files (references/*.md and scripts/*.py all exist and are named in the body); not a 5 because schemas are described inline rather than linked via markdown references to paper-schema.md/bundle-schema.md, leaving minor navigation gaps. | 4 / 5 |
Total | 15 / 20 Passed |