Content
62%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 delivers an unusually concrete, well-sequenced methodology — literal search templates, quantified scoring thresholds, probability mapping, risk rules, and a self-check checklist — but it is a 570-line monolith with no progressive disclosure: scoring rubrics, report template, glossary, and chart specs are all inlined, and a glossary of concepts Claude already knows pads the token cost. It would score near the top after splitting into reference files and trimming known-concept explanations.
Suggestions
Split the five-dimension scoring rubrics, report template, visualization specs, glossary, and special-case rules into separate files under references/ (e.g., references/pace-scoring.md, references/report-template.md, references/glossary.md) and keep SKILL.md as a concise overview linking to them one level deep.
Trim or move the 专业术语词典 — entries like T+1, 涨停板, MACD, KDJ, and 均线 explain concepts Claude already knows; keep only the instruction that report output must annotate terms on first use.
Replace the placeholder 使用示例 ("P=XX/20") with one fully worked example using real numbers (e.g., a completed 贵州茅台 scoring pass, total score, scenario probabilities, E(r), and risk parameters) to make the end-to-end process copy-paste concrete.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | At ~570 lines the body is noticeably verbose: it includes a glossary of concepts Claude already knows (T+1, 涨停板, MACD, KDJ, 均线, 融资余额), editorial commentary ("A股最核心的量化信号", "A股最重要的均线"), and elaborate ASCII-art boxes with {placeholders} that could be far terser. The genuinely novel PACE scoring material is diluted by this padding, matching the 'noticeably verbose; several unnecessary explanations or padded sections' anchor rather than the mostly-efficient score-3 anchor. | 2 / 5 |
Actionability | Highly concrete for an instruction-only skill: ten literal search-query templates, numeric scoring thresholds ("北向单日净买>10亿 → 强烈利多（8分）"), a score-to-probability conversion table, an E(r) formula with explicit decision rules ("若 E(r) > 0.5% 且 P_A/P_C > 1.5…"), a full report skeleton, and an exact file-naming convention. It falls short of 5 only because the worked example ("使用示例") is placeholder text ("P=XX/20") rather than a concrete end-to-end case with real numbers. | 4 / 5 |
Workflow Clarity | The six-step 执行流程 is clearly sequenced (parse input → 8+ parallel searches → per-dimension scoring with mandatory stated reasoning → score-to-probability/scenario mapping → risk parameters → report output), backed by explicit validation: a 12-item pre-output 质量检查清单, a data-completeness feedback path ("标注缺失 → 降低置信度 → 该维度取中间分"), and consistency checks ("三情景概率合计 = 100%"). This matches the anchor for clear sequencing with explicit validation steps, feedback loops, and checklists; this is an analysis skill, so the destructive/batch cap does not apply. | 5 / 5 |
Progressive Disclosure | There are no bundle files (no references/, scripts/, or assets/ exist) and no external references at all — the entire five-dimension scoring rubric, report template, visualization specs, glossary, and special-case rules are inlined in one 570-line SKILL.md. The many well-labeled ## sections keep it navigable (above the minimal-structure score-2 anchor), but content that clearly belongs in separate reference files (scoring rubrics, report template, glossary) is fully inline, so it cannot score 4 or 5. | 3 / 5 |
Total | 14 / 20 Passed |