Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →用于公告结构拆分的公告版式解析原子 skill，适用于通用行业文档解析场景。
.claude/skills/aifinlab-announcement-layout-parsing/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 35% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 1762% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 20% | 0% |
本 Skill 支持多种公告文档输入格式，核心数据来源包括：
> 说明：本 Skill 不包含文档采集功能，需要用户提供公告文档文件。建议文档格式规范，以便进行准确的版式解析。
本 Skill 提供全面的公告版式解析能力，涵盖多种解析功能：
json{ "document_info": { "filename": "announcement.pdf", "file_size": 512000, "page_count": 10, "language": "zh-CN", "announcement_type": "重大事项公告" }, "header": { "title": "关于重大资产重组的公告", "announcement_number": "2024-001", "company_name": "示例股份有限公司", "stock_code": "000001", "publish_date": "2024-03-15", "publish_org": "上海证券交易所" }, "structure": { "sections": [ { "level": 1, "title": "一、交易概述", "content": "交易概述内容...", "page": 1 }, { "level": 1, "title": "二、交易对方基本情况", "content": "交易对方基本情况...", "page": 2 } ] }, "tables": [ { "table_id": 1, "type": "financial_table", "position": { "page": 5, "section": "三、交易标的基本情况" }, "rows": 8, "columns": 4, "data": [ ["项目", "2024年", "2023年", "2022年"], ["营业收入", "1000", "900", "800"] ] } ], "key_information": { "company_name": "示例股份有限公司", "stock_code": "000001", "transaction_amount": 500000000, "transaction_date": "2024-03-20", "related_parties": ["关联公司A", "关联公司B"] } }
LICENSE 文件requirements.txt 为准| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 22,084 | 16,516 | -25% | 1 | 1 | 0% | 4,993 | 4,408 | -12% | 0 | 0 | — |
case-02 | fail→pass | 13,699 | 12,237 | -11% | 1 | 1 | 0% | 2,995 | 4,042 | +35% | 0 | 0 | — |
case-03 | fail→pass | 9,302 | 36,903 | +297% | 1 | 1 | 0% | 453 | 8,435 | +1762% | 0 | 0 | — |
case-04 | fail→pass | 10,118 | 5,933 | -41% | 1 | 1 | 0% | 1,725 | 2,483 | +44% | 0 | 0 | — |
case-05 | pass→pass | 7,742 | 4,541 | -41% | 1 | 1 | 0% | 1,176 | 2,201 | +87% | 0 | 0 | — |
case-06 | fail→fail | 16,782 | 25,573 | +52% | 1 | 1 | 0% | 2,082 | 5,786 | +178% | 0 | 0 | — |
case-07 | pass→pass | 13,382 | 10,625 | -21% | 1 | 1 | 0% | 1,954 | 2,973 | +52% | 0 | 0 | — |
case-08 | fail→pass | 10,886 | 4,898 | -55% | 1 | 1 | 0% | 2,014 | 2,411 | +20% | 0 | 0 | — |
case-09 | pass→pass | 19,039 | 8,568 | -55% | 1 | 1 | 0% | 3,029 | 3,144 | +4% | 0 | 0 | — |
case-10 | fail→pass | 7,288 | 10,679 | +47% | 1 | 1 | 0% | 1,205 | 3,045 | +153% | 0 | 0 | — |
case-11 | fail→pass | 4,001 | 10,293 | +157% | 1 | 1 | 0% | 614 | 3,051 | +397% | 0 | 0 | — |
case-12 | fail→pass | 13,292 | 7,029 | -47% | 1 | 1 | 0% | 2,105 | 2,513 | +19% | 0 | 0 | — |
case-13 | fail→pass | 12,600 | 6,828 | -46% | 1 | 1 | 0% | 2,228 | 2,795 | +25% | 0 | 0 | — |
case-14 | fail→pass | 12,823 | 6,645 | -48% | 1 | 1 | 0% | 2,351 | 2,714 | +15% | 0 | 0 | — |
case-15 | pass→pass | 15,677 | 15,056 | -4% | 1 | 1 | 0% | 2,456 | 3,410 | +39% | 0 | 0 | — |
case-16 | pass→fail | 9,507 | 4,628 | -51% | 1 | 1 | 0% | 1,714 | 2,139 | +25% | 0 | 0 | — |
case-17 | pass→pass | 15,183 | 11,579 | -24% | 1 | 1 | 0% | 2,018 | 3,227 | +60% | 0 | 0 | — |
case-18 | pass→pass | 22,248 | 10,874 | -51% | 1 | 1 | 0% | 1,691 | 3,034 | +79% | 0 | 0 | — |
case-19 | fail→pass | 5,774 | 14,849 | +157% | 1 | 1 | 0% | 493 | 4,416 | +796% | 0 | 0 | — |
case-20 | fail→pass | 6,559 | 11,753 | +79% | 1 | 1 | 0% | 310 | 3,102 | +901% | 0 | 0 | — |
case-21 | fail→pass | 20,150 | 5,356 | -73% | 1 | 1 | 0% | 2,746 | 2,380 | -13% | 0 | 0 | — |
case-22 | pass→pass | 13,970 | 12,482 | -11% | 1 | 1 | 0% | 2,066 | 3,080 | +49% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +55 percentage points is the difference between those two pass rates over the 20 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.