Install
openclaw skills install financial-report-analyzer-proEN: Extract, normalize and analyze key metrics from corporate financial reports (10-K, 10-Q, A-share annual/interim reports, IFRS/GAAP/CAS PDFs). Produces side-by-side YoY/QoQ comparisons, ratio analysis (ROE/ROA/gross margin/debt ratio/cash conversion cycle), and red-flag detection (earnings management signals, related-party tx anomalies, goodwill risk). Use when user uploads or names a financial report and asks "解读 / 分析 / 拆解 / 找风险点 / 对比同行 / extract financials". 中文:从企业财报(10-K、10-Q、A股年报半年报、IFRS/GAAP/CAS PDF)中抽取并归一化关键指标,输出同比/环比对比、关键比率分析(ROE/ROA/毛利率/资产负债率/现金转换周期),以及红旗信号检测(盈余管理迹象、关联交易异常、商誉风险)。当用户上传或提及财报并要求"解读/分析/拆解/找风险点/同行对比"时触发。
openclaw skills install financial-report-analyzer-proPull every metric that matters from a 200-page annual report in under 60 seconds, with traceable line-item provenance, peer benchmarking, and red-flag detection.
60 秒从 200 页年报抽出全部关键指标,每个数字可溯源到原文,附同行对标与风险信号识别。
Trigger keywords (中文): 分析财报、解读年报、半年报解读、利润表分析、资产负债表、现金流量表、ROE 分析、毛利率、商誉风险、关联交易、同行对比、行业对标、财报对比、招股说明书
Trigger keywords (EN): parse 10-K, analyze annual report, extract financials, ratio analysis, peer comparison, red flag detection, earnings quality
Supported inputs:
Do NOT use when:
The skill always produces a 5-section report:
| 段落 / Section | 内容 / Content |
|---|---|
| 1. 摘要 Executive Summary | 主营业务、规模、增长、盈利能力 5 行总结 |
| 2. 三表关键指标 Three-Statement KPIs | Revenue / Net Income / EBITDA / Total Assets / Equity / OCF |
| 3. 比率分析 Ratio Analysis | 盈利能力 / 偿债能力 / 营运能力 / 现金质量 / 杜邦分解 |
| 4. 红旗检测 Red Flags | 应收账款增速 vs 营收、商誉占净资产、关联交易、审计意见 |
| 5. 同行对标 Peer Benchmark | 行业中位数、分位数、相对排名 |
python3 scripts/ingest.py --input <pdf-or-xbrl> --out /tmp/raw_extract.json
pdfplumber)scripts/normalize_statements.py maps raw line items to a standardized chart of accounts (knowledge/coa_master.csv), handling:
python3 scripts/compute_ratios.py --input normalized.json --out ratios.json
Computes 40+ standard ratios documented in knowledge/ratio_definitions.md, with formulas matching CFA Institute conventions and CSRC disclosure requirements.
scripts/detect_red_flags.py runs 18 heuristic checks, including:
Each flag is documented with severity (🟢🟡🔴), evidence span, and recommended follow-up question.
If FINREPORT_PEER_DB is set (or built-in industry medians available), produces percentile ranks. Otherwise outputs absolute values with industry context notes.
python3 scripts/render_report.py --analysis ratios.json --flags flags.json --format md|json|html
For web channels with web-prism-artifact available, the skill automatically emits a structured artifact for inline rendering.
JSON structure:
{
"company": { "name": "...", "ticker": "...", "industry": "..." },
"period": { "fiscal_year": 2024, "reporting_basis": "CAS" },
"executive_summary": "...",
"kpis": { "revenue": {...}, "net_income": {...}, ... },
"ratios": { "roe": {...}, "gross_margin": {...}, ... },
"red_flags": [ { "code": "RF03", "severity": "🟡", "title": "...", "evidence": "...", "page": 47 } ],
"peer_benchmark": { "industry": "...", "rankings": {...} },
"provenance": { /* page+line offset for every number */ }
}
For human consumption, --format md produces a Markdown report with embedded tables and red-flag callouts.
{page, table, row} pointing back to the source document.<output>.audit.log for compliance.本技能仅做描述性分析,绝不输出买卖评级或股价预测;所有数字可溯源;缺失数据明确标注不做估算;红旗信号仅作为"需进一步核查"的提示,不作定性结论。
python3 scripts/run_pipeline.py --input 600519_2024_annual.pdf --output report.json
python3 scripts/render_report.py --input report.json --format md > 茅台2024分析.md
python3 scripts/compare_periods.py \
--current 2024_annual.pdf \
--prior 2023_annual.pdf \
--output yoy_comparison.md
python3 scripts/peer_benchmark.py \
--target 600519_2024_annual.pdf \
--peers 000858,000568,000799 \
--output peer_report.md
python3 scripts/run_pipeline.py --input report.pdf --mode red-flag-only --output flags.json
cd tests && python3 -m pytest -v
Test fixtures include:
finance accounting 10-K annual-report ratio-analysis red-flag IFRS GAAP CAS 财报 年报 财务分析