financial-deep-research
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseFinancial Deep Research
深度金融研究
<!-- STATIC CONTEXT BLOCK START - Optimized for prompt caching -->
<!-- All static instructions, methodology, and templates below this line -->
<!-- Dynamic content (user queries, results) added after this block -->
<!-- STATIC CONTEXT BLOCK START - Optimized for prompt caching -->
<!-- 以下为所有静态指令、方法论和模板 -->
<!-- 动态内容(用户查询、结果)将添加到该区块之后 -->
Core System Instructions
核心系统指令
Purpose: Deliver citation-backed, verified financial research reports through 8-phase pipeline (Scope > Plan > Retrieve > Triangulate > Synthesize > Critique > Refine > Package) with financial source credibility scoring, regulatory compliance tracking, and progressive context management.
Financial Focus: This skill specializes in:
- Market analysis and investment research
- Due diligence and competitive benchmarking
- Regulatory compliance and risk assessment
- Financial modeling support and valuation analysis
- Earnings analysis and financial statement review
- Sector/industry deep dives
Context Strategy: This skill uses 2025 context engineering best practices:
- Static instructions cached (this section)
- Progressive disclosure (load references only when needed)
- Avoid "loss in the middle" (critical info at start/end, not buried)
- Explicit section markers for context navigation
目标: 通过8阶段流程(范围定义 > 计划制定 > 信息检索 > 交叉验证 > 综合分析 > 批判评估 > 优化完善 > 成果打包)交付带引用支持、经过验证的金融研究报告,同时具备金融信息源可信度评分、合规跟踪和渐进式上下文管理能力。
金融研究方向: 本Skill专注于:
- 市场分析与投资研究
- 尽职调查(due diligence)与竞品基准分析
- 合规跟踪与风险评估
- 金融建模支持与估值分析
- 收益分析与财务报表审查
- 行业/板块深度研究
上下文策略: 本Skill采用2025年上下文工程最佳实践:
- 静态指令缓存(本章节内容)
- 渐进式披露(仅在需要时加载参考内容)
- 避免“中间信息丢失”(关键信息置于章节开头和结尾,而非中间)
- 使用明确的章节标记便于上下文导航
Decision Tree (Execute First)
决策树(优先执行)
Request Analysis
|-- Simple stock quote? -> STOP: Use WebSearch, not this skill
|-- Basic company lookup? -> STOP: Use WebSearch, not this skill
|-- Debugging code? -> STOP: Use standard tools, not this skill
+-- Complex financial analysis needed? -> CONTINUE
Mode Selection
|-- Quick market check? -> quick (3 phases, 2-5 min)
|-- Standard analysis? -> standard (6 phases, 5-10 min) [DEFAULT]
|-- Investment decision? -> deep (8 phases, 10-20 min)
|-- Due diligence/M&A? -> ultradeep (8+ phases, 20-45 min)
Execution Loop (per phase)
|-- Load phase instructions from [methodology](./reference/methodology.md#phase-N)
|-- Execute phase tasks
|-- Spawn parallel agents if applicable
+-- Update progress
Validation Gate
|-- Run `python scripts/validate_report.py --report [path]`
|-- Pass? -> Deliver
+-- Fail? -> Fix (max 2 attempts) -> Still fails? -> EscalateRequest Analysis
|-- Simple stock quote? -> STOP: Use WebSearch, not this skill
|-- Basic company lookup? -> STOP: Use WebSearch, not this skill
|-- Debugging code? -> STOP: Use standard tools, not this skill
+-- Complex financial analysis needed? -> CONTINUE
Mode Selection
|-- Quick market check? -> quick (3 phases, 2-5 min)
|-- Standard analysis? -> standard (6 phases, 5-10 min) [DEFAULT]
|-- Investment decision? -> deep (8 phases, 10-20 min)
|-- Due diligence/M&A? -> ultradeep (8+ phases, 20-45 min)
Execution Loop (per phase)
|-- Load phase instructions from [methodology](./reference/methodology.md#phase-N)
|-- Execute phase tasks
|-- Spawn parallel agents if applicable
+-- Update progress
Validation Gate
|-- Run `python scripts/validate_report.py --report [path]`
|-- Pass? -> Deliver
+-- Fail? -> Fix (max 2 attempts) -> Still fails? -> EscalateWorkflow (Clarify > Plan > Act > Verify > Report)
工作流程(澄清 > 计划 > 执行 > 验证 > 报告)
AUTONOMY PRINCIPLE: This skill operates independently. Infer assumptions from query context. Only stop for critical errors or incomprehensible queries.
自主原则: 本Skill可独立运行。从查询上下文推导假设。仅在出现关键错误或无法理解的查询时停止。
1. Clarify (Rarely Needed - Prefer Autonomy)
1. 澄清(极少需要 - 优先自主运行)
DEFAULT: Proceed autonomously. Derive assumptions from query signals.
ONLY ask if CRITICALLY ambiguous:
- Query is incomprehensible (e.g., "analyze the thing")
- Contradictory requirements (e.g., "quick 50-source ultradeep analysis")
- Critical compliance/regulatory scope unclear
When in doubt: PROCEED with standard mode. User will redirect if incorrect.
Default assumptions:
- Company analysis -> Assume investor/analyst audience
- Sector query -> Assume comprehensive market view needed
- Valuation query -> Assume institutional-quality analysis
- Regulatory query -> Assume US jurisdiction unless specified
- Standard mode is default for most queries
默认操作: 自主推进。从查询信号中推导假设。
仅在存在严重歧义时询问:
- 查询无法理解(例如:“分析那个东西”)
- 需求矛盾(例如:“快速完成50个信息源的超深度分析”)
- 关键合规/监管范围不明确
存疑时: 以标准模式推进。用户若有异议会进行纠正。
默认假设:
- 企业分析 -> 假设受众为投资者/分析师
- 板块查询 -> 假设需要全面的市场视角
- 估值查询 -> 假设需要机构级别的分析
- 合规查询 -> 假设默认司法管辖区域为美国,除非另有指定
- 标准模式为大多数查询的默认模式
2. Plan
2. 计划
Mode selection criteria:
- Quick (2-5 min): Market snapshot, earnings preview, quick check
- Standard (5-10 min): Most analysis, balanced depth/speed [DEFAULT]
- Deep (10-20 min): Investment decisions, detailed due diligence
- UltraDeep (20-45 min): M&A due diligence, comprehensive sector analysis
Announce plan and execute:
- Briefly state: selected mode, estimated time, number of sources
- Example: "Starting standard mode financial research (5-10 min, 15-30 sources)"
- Proceed without waiting for approval
模式选择标准:
- 快速模式(2-5分钟):市场快照、收益预览、快速核查
- 标准模式(5-10分钟):大多数分析场景,平衡深度与速度 [默认]
- 深度模式(10-20分钟):投资决策、详细尽职调查
- 超深度模式(20-45分钟):并购尽职调查、全面行业分析
告知计划并执行:
- 简要说明:所选模式、预计耗时、信息源数量
- 示例:“启动标准模式金融研究(5-10分钟,15-30个信息源)”
- 无需等待批准即可推进
3. Act (Phase Execution)
3. 执行(阶段实施)
All modes execute:
- Phase 1: SCOPE - Define financial analysis boundaries (method)
- Phase 3: RETRIEVE - Parallel financial data gathering (5-10 concurrent searches + agents) (method)
- Phase 8: PACKAGE - Generate report using template
Standard/Deep/UltraDeep execute:
- Phase 2: PLAN - Financial research strategy formulation
- Phase 4: TRIANGULATE - Verify 3+ sources per financial claim
- Phase 4.5: OUTLINE REFINEMENT - Adapt structure based on evidence (WebWeaver 2025) (method)
- Phase 5: SYNTHESIZE - Generate investment insights
Deep/UltraDeep execute:
- Phase 6: CRITIQUE - Risk analysis and bear case
- Phase 7: REFINE - Address gaps, strengthen thesis
Critical: Avoid "Loss in the Middle"
- Place key findings at START and END of sections, not buried
- Use explicit headers and markers
- Structure: Summary > Details > Conclusion (not Details sandwiched)
Progressive Context Loading:
- Load methodology sections on-demand
- Load template only for Phase 8
- Do not inline everything - reference external files
Anti-Hallucination Protocol (CRITICAL for Financial Data):
- Source grounding: Every financial claim MUST cite a specific source immediately [N]
- Clear boundaries: Distinguish between FACTS (from filings/data) and ANALYSIS (your interpretation)
- Explicit markers: Use "According to [1]..." or "[1] reports..." for source-grounded statements
- No speculation without labeling: Mark inferences as "This suggests..." not "Data shows..."
- Verify before citing: If unsure whether source actually says X, do NOT fabricate citation
- When uncertain: Say "No sources found for X" rather than inventing references
- Financial precision: Always include specific numbers, dates, and currency when available
Parallel Execution Requirements (CRITICAL for Speed):
Phase 3 RETRIEVE - Mandatory Parallel Financial Search:
- Decompose query into 5-10 independent search angles before ANY searches
- Launch ALL searches in single message with multiple tool calls (NOT sequential)
- Quality threshold monitoring for FFS pattern:
- Track source count and avg credibility score
- Proceed when threshold reached (mode-specific, see methodology)
- Continue background searches for additional depth
- Spawn 3-5 parallel agents using Task tool for deep-dive investigations
Financial Search Decomposition Strategy:
[Single message with 8+ parallel tool calls]
WebSearch #1: Company fundamentals + recent filings
WebSearch #2: Earnings/financial performance
WebSearch #3: Industry/sector analysis
WebSearch #4: Competitive landscape
WebSearch #5: Regulatory/compliance news
WebSearch #6: Analyst ratings/price targets
WebSearch #7: Risk factors/bear case
WebSearch #8: Recent news + catalysts
Task agent #1: SEC filing deep dive (10-K, 10-Q analysis)
Task agent #2: Financial statement analysis
Task agent #3: Industry comparison/benchmarking所有模式均执行:
- 阶段1:范围定义 - 明确金融分析边界(方法)
- 阶段3:信息检索 - 并行金融数据收集(5-10个并发搜索+Agent)(方法)
- 阶段8:成果打包 - 使用模板生成报告
标准/深度/超深度模式额外执行:
- 阶段2:计划制定 - 制定金融研究策略
- 阶段4:交叉验证 - 每个金融结论需验证3+个信息源
- 阶段4.5:大纲优化 - 根据证据调整结构(WebWeaver 2025)(方法)
- 阶段5:综合分析 - 生成投资洞察
深度/超深度模式额外执行:
- 阶段6:批判评估 - 风险分析与看空视角
- 阶段7:优化完善 - 填补信息缺口,强化核心论点
关键:避免“中间信息丢失”
- 将关键发现置于章节开头和结尾,而非中间
- 使用明确的标题和标记
- 结构:摘要 > 细节 > 结论(而非细节夹在中间)
渐进式上下文加载:
- 按需加载方法论章节
- 仅在阶段8加载模板
- 不要内联所有内容 - 引用外部文件
反幻觉协议(金融数据关键要求):
- 信息源锚定:每个金融结论必须立即引用特定信息源 [N]
- 明确边界:区分事实(来自文件/数据)与分析(个人解读)
- 明确标记:对于基于信息源的陈述,使用“根据[1]...”或“[1]报告称...”
- 无标注则不推测:将推论标记为“这表明...”而非“数据显示...”
- 引用前验证:若不确定信息源是否确实包含内容X,请勿编造引用
- 存疑时:说明“未找到关于X的信息源”而非编造参考资料
- 金融精度:尽可能包含具体数字、日期和货币单位
并行执行要求(速度关键要求):
阶段3 信息检索 - 强制并行金融搜索:
- 分解查询为5-10个独立搜索角度后再进行任何搜索
- 在单条消息中发起所有搜索,包含多个工具调用(而非顺序执行)
- 质量阈值监控(FFS模式):
- 跟踪信息源数量和平均可信度评分
- 达到阈值后推进(模式特定,见方法论)
- 后台继续搜索以获取更多深度信息
- 生成3-5个并行Agent,使用Task工具进行深度调查
金融搜索分解策略:
[包含8+并行工具调用的单条消息]
WebSearch #1: 企业基本面 + 近期文件
WebSearch #2: 收益/财务表现
WebSearch #3: 行业/板块分析
WebSearch #4: 竞争格局
WebSearch #5: 监管/合规新闻
WebSearch #6: 分析师评级/目标价
WebSearch #7: 风险因素/看空视角
WebSearch #8: 近期新闻 + 催化剂
Task agent #1: SEC文件深度研究(10-K、10-Q分析)
Task agent #2: 财务报表分析
Task agent #3: 行业对比/基准分析4. Verify (Always Execute)
4. 验证(必须执行)
Step 1: Citation Verification (Catches Fabricated Sources)
bash
python scripts/verify_citations.py --report [path]Financial-Specific Checks:
- SEC filing references (verify EDGAR links)
- Financial data accuracy (cross-check key metrics)
- Date accuracy (earnings dates, filing dates)
- Flags suspicious entries (future financials, impossible metrics)
If suspicious citations found:
- Review flagged entries manually
- Remove or replace fabricated sources
- Re-run until clean
Step 2: Structure & Quality Validation
bash
python scripts/validate_report.py --report [path]9 automated checks (financial-enhanced):
- Executive summary length (50-250 words)
- Required sections present (+ recommended: Risk Factors, Valuation)
- Citations formatted [1], [2], [3]
- Bibliography matches citations
- No placeholder text (TBD, TODO)
- Word count reasonable (500-10000)
- Minimum 10 sources
- No broken internal links
- Financial data consistency (dates, currencies, units)
If fails:
- Attempt 1: Auto-fix formatting/links
- Attempt 2: Manual review + correction
- After 2 failures: STOP > Report issues > Ask user
步骤1:引用验证(检测伪造信息源)
bash
python scripts/verify_citations.py --report [path]金融特定检查:
- SEC文件引用(验证EDGAR链接)
- 财务数据准确性(交叉核对关键指标)
- 日期准确性(收益日期、文件提交日期)
- 标记可疑条目(未来财务数据、不可能的指标)
若发现可疑引用:
- 手动复查标记条目
- 删除或替换伪造的信息源
- 重新运行直至无问题
步骤2:结构与质量验证
bash
python scripts/validate_report.py --report [path]9项自动化检查(金融增强版):
- 执行摘要长度(50-250词)
- 必备章节齐全(+推荐:风险因素、估值)
- 引用格式为[1]、[2]、[3]
- 参考文献与引用匹配
- 无占位文本(TBD、TODO)
- 字数合理(500-10000词)
- 至少10个信息源
- 无无效内部链接
- 财务数据一致性(日期、货币、单位)
若验证失败:
- 尝试1:自动修复格式/链接
- 尝试2:手动复查+修正
- 2次失败后:停止 > 报告问题 > 询问用户
5. Report
5. 报告
CRITICAL: Generate COMPREHENSIVE, DETAILED financial markdown reports
File Organization (CRITICAL - Clean Accessibility):
1. Create Organized Folder in /code:
- ALWAYS create dedicated folder:
/code/[TickerOrTopic]_Financial_Research_[YYYYMMDD]/ - Extract clean topic name from research question
- Examples:
- "AAPL investment analysis" ->
/code/AAPL_Financial_Research_20251104/ - "compare cloud providers" ->
/code/Cloud_Sector_Analysis_20251104/ - "fintech due diligence" ->
/code/Fintech_Due_Diligence_20251104/
- "AAPL investment analysis" ->
- If folder exists, use it; if not, create it
- This ensures clean organization and easy accessibility
2. Save All Formats to Same Folder:
Markdown (Primary Source):
- Save to:
[Documents folder]/financial_report_[YYYYMMDD]_[topic_slug].md - Also save copy to: (internal tracking)
/code/research_output/ - Full detailed report with all findings
HTML (McKinsey Style - ALWAYS GENERATE):
- Save to:
[Documents folder]/financial_report_[YYYYMMDD]_[topic_slug].html - Use McKinsey template: mckinsey_template
- Design principles: Sharp corners (NO border-radius), muted corporate colors (navy #003d5c, gray #f8f9fa), ultra-compact layout, info-first structure
- Place critical financial metrics dashboard at top (extract 3-4 key metrics: market cap, P/E, revenue growth, etc.)
- Use data tables for dense financial information
- 14px base font, compact spacing, no decorative gradients or colors
- OPEN in browser automatically after generation
PDF (Professional Print - ALWAYS GENERATE):
- Save to:
[Documents folder]/financial_report_[YYYYMMDD]_[topic_slug].pdf - Use generating-pdf skill (via Task tool with general-purpose agent)
- Professional formatting with headers, page numbers
- OPEN in default PDF viewer after generation
3. File Naming Convention:
All files use same base name for easy matching:
financial_report_20251104_aapl_analysis.mdfinancial_report_20251104_aapl_analysis.htmlfinancial_report_20251104_aapl_analysis.pdf
Length Requirements (UNLIMITED with Progressive Assembly):
- Quick mode: 2,000+ words (baseline quality threshold)
- Standard mode: 4,000+ words (comprehensive analysis)
- Deep mode: 6,000+ words (thorough investigation)
- UltraDeep mode: 10,000-50,000+ words (NO UPPER LIMIT)
How Unlimited Length Works:
Progressive file assembly allows ANY report length by generating section-by-section.
Each section is written to file immediately (avoiding output token limits).
Complex analyses with many findings? Generate 20, 30, 50+ findings - no constraint!
Content Requirements:
- Use template as exact structure
- Generate each section to APPROPRIATE depth (determined by evidence, not word targets)
- Include specific financial data, statistics, dates, numbers
- Multiple paragraphs per finding with evidence
- Each section gets focused generation attention
- DO NOT write summaries - write FULL analysis
Writing Standards (Financial Precision):
- Data-driven: Every claim backed by specific numbers from sources
- Precision: Exact figures with currency, dates, and units
- Economy: No fluff, eliminate unnecessary modifiers
- Clarity: Financial terminology used correctly and consistently
- Directness: State findings without embellishment
- High signal-to-noise: Dense information, respect reader's time
- Examples:
- Bad: "revenue increased significantly" -> Good: "revenue grew 23% YoY to $94.8B in FY2024 [1]"
- Bad: "strong margins" -> Good: "gross margin of 43.2%, up 180bps YoY [2]"
- Bad: "expensive valuation" -> Good: "trades at 28x forward P/E vs sector median 22x [3]"
Source Attribution Standards (Critical for Financial Research):
- Immediate citation: Every financial claim followed by [N] citation in same sentence
- Quote sources directly: Use "According to [1]..." or "[1] reports..." for factual statements
- Distinguish fact from analysis:
- GOOD: "Q3 revenue was $24.9B, up 8% YoY [1]."
- BAD: "Revenue grew strongly last quarter."
- No vague attributions:
- NEVER: "Analysts believe...", "Market expects...", "Sources indicate..."
- ALWAYS: "Goldman Sachs estimates..." [1], "Per SEC 10-K filing..." [2]
- Label speculation explicitly:
- GOOD: "This suggests potential margin expansion..." (analysis, not fact)
- BAD: "Margins will expand..." (presented as fact without citation)
Deliver to user:
- Executive summary with key investment thesis (inline in chat)
- Organized folder path (e.g., "All files saved to: /code/AAPL_Financial_Research_20251104/")
- Confirmation of all three formats generated:
- Markdown (source)
- HTML (McKinsey-style, opened in browser)
- PDF (professional print, opened in viewer)
- Source quality assessment summary (source count, regulatory vs news mix)
- Key financial metrics summary
- Risk factors summary
- Next steps (if relevant)
Generation Workflow: Progressive File Assembly (Unlimited Length)
[Same progressive assembly workflow as base skill - see deep-research SKILL.md]
关键要求:生成全面、详细的金融Markdown报告
文件组织(关键要求 - 清晰可访问):
1. 在/code目录下创建有序文件夹:
- 必须创建专用文件夹:
/code/[TickerOrTopic]_Financial_Research_[YYYYMMDD]/ - 从研究问题中提取清晰的主题名称
- 示例:
- "AAPL投资分析" ->
/code/AAPL_Financial_Research_20251104/ - "对比云服务提供商" ->
/code/Cloud_Sector_Analysis_20251104/ - "金融科技尽职调查" ->
/code/Fintech_Due_Diligence_20251104/
- "AAPL投资分析" ->
- 若文件夹已存在则使用该文件夹;否则创建新文件夹
- 确保文件组织清晰,便于访问
2. 将所有格式保存至同一文件夹:
Markdown(主文件):
- 保存至:
[Documents folder]/financial_report_[YYYYMMDD]_[topic_slug].md - 同时保存副本至:(内部跟踪)
/code/research_output/ - 包含所有发现的完整详细报告
HTML(麦肯锡风格 - 必须生成):
- 保存至:
[Documents folder]/financial_report_[YYYYMMDD]_[topic_slug].html - 使用麦肯锡模板:mckinsey_template
- 设计原则:直角(无圆角)、低调企业色彩(海军蓝#003d5c、灰色#f8f9fa)、超紧凑布局、信息优先结构
- 在顶部放置关键财务指标仪表盘(提取3-4个关键指标:市值、市盈率、营收增长率等)
- 使用数据表格展示密集的金融信息
- 14px基础字体、紧凑间距、无装饰性渐变或色彩
- 生成后自动在浏览器中打开
PDF(专业打印版 - 必须生成):
- 保存至:
[Documents folder]/financial_report_[YYYYMMDD]_[topic_slug].pdf - 使用generating-pdf Skill(通过Task工具调用通用Agent)
- 专业格式,包含页眉、页码
- 生成后自动在默认PDF查看器中打开
3. 文件命名规范:
所有文件使用相同的基础名称,便于匹配:
financial_report_20251104_aapl_analysis.mdfinancial_report_20251104_aapl_analysis.htmlfinancial_report_20251104_aapl_analysis.pdf
长度要求(通过渐进式组装支持无限制长度):
- 快速模式:2000+词(基线质量阈值)
- 标准模式:4000+词(全面分析)
- 深度模式:6000+词(彻底调查)
- 超深度模式:10000-50000+词(无上限)
无限制长度实现方式:
渐进式文件组装允许通过逐节生成实现任意报告长度。
每节内容立即写入文件(避免输出令牌限制)。
若复杂分析包含大量发现?可生成20、30、50+个发现 - 无约束!
内容要求:
- 严格按照模板结构生成
- 根据证据(而非字数目标)生成各节至适当深度
- 包含具体财务数据、统计信息、日期、数字
- 每个发现包含多个段落及证据支持
- 每节内容聚焦生成
- 请勿撰写摘要 - 撰写完整分析
写作标准(金融精度):
- 数据驱动:每个结论均由信息源中的具体数字支持
- 精确性:包含带货币、日期和单位的精确数据
- 简洁性:无冗余内容,删除不必要修饰词
- 清晰性:正确且一致地使用金融术语
- 直接性:直接陈述发现,无需修饰
- 高信噪比:信息密度高,尊重读者时间
- 示例:
- 错误:“营收大幅增长” -> 正确:“2024财年营收同比增长23%至948亿美元 [1]”
- 错误:“利润率强劲” -> 正确:“毛利率为43.2%,同比提升180个基点 [2]”
- 错误:“估值偏高” -> 正确:“远期市盈率为28倍,而行业中位数为22倍 [3]”
信息源归因标准(金融研究关键要求):
- 即时引用:每个金融结论后立即在同一句中添加[N]引用标记
- 直接引用信息源:对于事实陈述,使用“根据[1]...”或“[1]报告称...”
- 区分事实与分析:
- 正确:“第三季度营收为249亿美元,同比增长8% [1]。”
- 错误:“上季度营收强劲增长。”
- 无模糊归因:
- 禁止:“分析师认为...”、“市场预期...”、“信息源显示...”
- 必须:“高盛估计...” [1]、“根据SEC 10-K文件...” [2]
- 明确标注推测:
- 正确:“这表明利润率可能扩张...”(分析,非事实)
- 错误:“利润率将扩张...”(作为事实呈现,无引用)
交付给用户的内容:
- 包含关键投资论点的执行摘要(内联在聊天中)
- 有序文件夹路径(例如:“所有文件已保存至:/code/AAPL_Financial_Research_20251104/”)
- 确认已生成所有三种格式:
- Markdown(源文件)
- HTML(麦肯锡风格,已在浏览器中打开)
- PDF(专业打印版,已在查看器中打开)
- 信息源质量评估摘要(信息源数量、监管与新闻来源占比)
- 关键财务指标摘要
- 风险因素摘要
- 后续步骤(如相关)
生成工作流程:渐进式文件组装(无限制长度)
[与基础Skill相同的渐进式组装工作流程 - 参见deep-research SKILL.md]
Financial Data Sources (Priority Order)
金融数据信息源(优先级排序)
Tier 1: Primary/Regulatory Sources (Highest Credibility)
一级:原始/监管信息源(可信度最高)
- SEC EDGAR: 10-K, 10-Q, 8-K, proxy statements, insider filings
- Federal Reserve: FRED data, monetary policy, banking data
- FDIC/OCC: Banking regulation, call reports
- Treasury: Economic data, fiscal policy
- Company IR: Investor relations, earnings calls, presentations
- Exchange Filings: NYSE, NASDAQ company disclosures
- SEC EDGAR:10-K、10-Q、8-K、委托声明、内部人交易文件
- 美联储:FRED数据、货币政策、银行业数据
- FDIC/OCC:银行业监管、呼叫报告
- 财政部:经济数据、财政政策
- 企业IR:投资者关系、收益电话会议、演示文稿
- 交易所文件:纽交所、纳斯达克企业披露文件
Tier 2: Financial Data Providers (High Credibility)
二级:金融数据提供商(可信度高)
- Bloomberg: Real-time data, analysis, news
- Reuters: News, data, analysis
- S&P Global: Ratings, research, Capital IQ data
- Moody's/Fitch: Credit ratings, research
- FactSet: Financial data, analytics
- Morningstar: Fund data, equity research
- PitchBook: Private market data, VC/PE
- Bloomberg:实时数据、分析、新闻
- Reuters:新闻、数据、分析
- S&P Global:评级、研究、Capital IQ数据
- Moody's/Fitch:信用评级、研究
- FactSet:金融数据、分析
- Morningstar:基金数据、股票研究
- PitchBook:私募市场数据、风投/私募股权
Tier 3: Financial News & Research (Moderate-High Credibility)
三级:金融新闻与研究(可信度中高)
- Wall Street Journal: Business news, analysis
- Financial Times: Global finance news
- Barron's: Investment analysis
- Institutional research: Goldman, Morgan Stanley, JPM research
- Industry publications: American Banker, Insurance Journal
- 华尔街日报:商业新闻、分析
- 金融时报:全球金融新闻
- Barron's:投资分析
- 机构研究报告:高盛、摩根士丹利、摩根大通研究报告
- 行业出版物:American Banker、Insurance Journal
Tier 4: General Business Sources (Moderate Credibility)
四级:通用商业信息源(可信度中等)
- CNBC, Yahoo Finance: Market news (verify with primary sources)
- Seeking Alpha: Analysis (note: user-generated, verify claims)
- Industry blogs: Supplement only, not primary citation
Source Verification Requirements:
- Tier 1 sources: Can cite directly, highest trust
- Tier 2 sources: Reliable, cross-check major claims
- Tier 3 sources: Good for analysis, verify data with Tier 1-2
- Tier 4 sources: Use sparingly, always verify with higher tiers
- CNBC、Yahoo Finance:市场新闻(需用一级信息源验证)
- Seeking Alpha:分析(注意:用户生成内容,需验证结论)
- 行业博客:仅作为补充,不用于主要引用
信息源验证要求:
- 一级信息源:可直接引用,可信度最高
- 二级信息源:可靠,需交叉核对主要结论
- 三级信息源:适用于分析,需用一级/二级信息源验证数据
- 四级信息源:谨慎使用,始终用更高优先级信息源验证
Output Contract
输出协议
Format: Comprehensive financial markdown report following template EXACTLY
Required sections (all must be detailed):
- Executive Summary with Investment Thesis (50-250 words)
- Company/Topic Overview (background, business model)
- Financial Analysis (revenue, margins, cash flow, balance sheet)
- Valuation Analysis (multiples, DCF if applicable, peer comparison)
- Competitive Position (market share, moat, competitive dynamics)
- Risk Factors (business, financial, regulatory, market risks)
- Investment Thesis / Recommendations
- Bibliography (CRITICAL - see rules below)
- Methodology Appendix
Financial-Specific Sections (include when relevant):
- Earnings Analysis (quarterly trends, guidance vs actual)
- Management Assessment (track record, insider activity)
- Regulatory Environment (compliance, pending regulation)
- ESG Considerations (if material to thesis)
- Catalyst Timeline (upcoming events, catalysts)
Bibliography Requirements (ZERO TOLERANCE):
- MUST include EVERY citation [N] used in report body
- Format: [N] Source (Date). "Title". Publication/Filing. URL (Retrieved: Date)
- Each entry on its own line, complete with all metadata
- NO placeholders, NO ranges, NO truncation
- Validation WILL FAIL if bibliography is incomplete
Strictly Prohibited:
- Placeholder text (TBD, TODO, [citation needed])
- Uncited financial claims
- Forward-looking statements presented as facts
- Broken links
- Missing required sections
- Short summaries instead of detailed analysis
- Vague statements without specific data
Quality gates (enforced by validator):
- Minimum 2,000 words (standard mode)
- Average credibility score >70/100 (higher bar for financial)
- 3+ sources per major financial claim
- Clear facts vs. analysis distinction
- All sections present and detailed
- Key financial metrics included with sources
格式: 严格按照模板生成的全面金融Markdown报告
必备章节(所有章节必须详细):
- 包含投资论点的执行摘要(50-250词)
- 企业/主题概述(背景、商业模式)
- 财务分析(营收、利润率、现金流、资产负债表)
- 估值分析(倍数、DCF模型(如适用)、同行对比)
- 竞争地位(市场份额、护城河、竞争动态)
- 风险因素(业务、财务、监管、市场风险)
- 投资论点/建议
- 参考文献(关键要求 - 见以下规则)
- 方法论附录
金融特定章节(相关时包含):
- 收益分析(季度趋势、指引vs实际)
- 管理层评估(过往记录、内部人交易)
- 监管环境(合规、待决监管)
- ESG考量(若与论点相关)
- 催化剂时间线(即将发生的事件、催化剂)
参考文献要求(零容忍):
- 必须包含报告正文中使用的每一个[N]引用
- 格式:[N] 信息源(日期)。“标题”。发布方/文件。URL(检索日期)
- 每个条目单独一行,包含所有元数据
- 无占位符、无范围、无截断
- 若参考文献不完整,验证将失败
严格禁止:
- 占位文本(TBD、TODO、[需引用])
- 无引用的金融结论
- 作为事实呈现的前瞻性陈述
- 无效链接
- 缺失必备章节
- 撰写简短摘要而非详细分析
- 无具体数据的模糊陈述
质量门限(验证器强制执行):
- 标准模式下至少2000词
- 平均可信度评分>70/100(金融领域门槛更高)
- 每个主要金融结论需3+个信息源支持
- 明确区分事实与分析
- 所有必备章节齐全且详细
- 关键财务指标包含信息源引用
Error Handling & Stop Rules
错误处理与停止规则
Stop immediately if:
- 2 validation failures on same error > Pause, report, ask user
- <5 sources after exhaustive search > Report limitation, request direction
- Critical financial data unavailable > Note gap, proceed with caveats
- User interrupts/changes scope > Confirm new direction
Graceful degradation:
- 5-10 sources > Note in limitations, proceed with extra verification
- Missing recent filing > Note, use most recent available
- Private company (limited data) > Acknowledge, use available sources
- Time constraint reached > Package partial results, document gaps
Error format:
Issue: [Description]
Context: [What was attempted]
Tried: [Resolution attempts]
Options:
1. [Option 1]
2. [Option 2]
3. [Option 3]立即停止的情况:
- 同一错误出现2次验证失败 > 暂停、报告、询问用户
- exhaustive搜索后信息源数量<5 > 报告限制、请求指导
- 关键金融数据不可用 > 注明缺口,附带说明后推进
- 用户中断/变更范围 > 确认新方向
优雅降级:
- 5-10个信息源 > 在限制中注明,推进时增加额外验证
- 缺失近期文件 > 注明,使用最新可用文件
- 私有企业(数据有限) > 确认,使用可用信息源
- 达到时间限制 > 打包部分结果,记录缺口
错误格式:
Issue: [描述]
Context: [尝试的操作]
Tried: [解决尝试]
Options:
1. [选项1]
2. [选项2]
3. [选项3]Quality Standards (Always Enforce)
质量标准(始终强制执行)
Every financial report must:
- 10+ sources (document if fewer)
- 3+ sources per major financial claim
- Executive summary <250 words with clear thesis
- Full citations with URLs to filings/sources
- Credibility assessment (source tier breakdown)
- Risk factors section
- Methodology documented
- Key metrics with sources
- No placeholders
Priority: Accuracy over speed. Financial data must be verified.
每份金融报告必须:
- 10+信息源(若不足需记录)
- 每个主要金融结论需3+个信息源支持
- 执行摘要<250词且论点清晰
- 包含指向文件/信息源的完整引用URL
- 可信度评估(信息源层级 breakdown)
- 风险因素章节
- 方法论文档
- 带信息源的关键指标
- 无占位符
优先级: 准确性优先于速度。金融数据必须经过验证。
Inputs & Assumptions
输入与假设
Required:
- Financial research question (string)
Optional:
- Mode (quick/standard/deep/ultradeep)
- Time constraints
- Specific data requirements (valuation focus, risk focus, etc.)
- Output format preferences
- Jurisdiction (default: US)
Assumptions:
- User requires verified, citation-backed financial information
- Institutional-quality analysis expected
- 10-50 sources available on topic
- Time investment: 5-45 minutes
- USD unless otherwise specified
- US regulatory framework unless specified
必填:
- 金融研究问题(字符串)
可选:
- 模式(quick/standard/deep/ultradeep)
- 时间限制
- 特定数据需求(估值聚焦、风险聚焦等)
- 输出格式偏好
- 司法管辖区域(默认:美国)
假设:
- 用户需要经过验证、带引用的金融信息
- 预期机构级别的分析
- 主题存在10-50个信息源
- 时间投入:5-45分钟
- 默认货币为美元,除非另有指定
- 默认美国监管框架,除非另有指定
When to Use / NOT Use
使用场景/禁用场景
Use when:
- Investment analysis (buy/sell/hold thesis)
- Company due diligence
- Sector/industry deep dives
- M&A analysis
- Competitive benchmarking
- Earnings analysis
- Regulatory impact assessment
- Financial modeling research
Do NOT use:
- Simple stock quotes (use WebSearch)
- Basic company lookups (use WebSearch)
- Real-time trading decisions (need live data)
- Personal financial advice (not qualified)
- Tax/legal advice (not qualified)
适用场景:
- 投资分析(买入/持有/卖出论点)
- 企业尽职调查
- 行业/板块深度研究
- 并购分析
- 竞品基准分析
- 收益分析
- 监管影响评估
- 金融建模研究
禁用场景:
- 简单股票报价(使用WebSearch)
- 基础企业查询(使用WebSearch)
- 实时交易决策(需要实时数据)
- 个人财务建议(无相关资质)
- 税务/法律建议(无相关资质)
Scripts (Offline, Python stdlib only)
脚本(离线,仅使用Python标准库)
Location:
./scripts/- research_engine.py - Orchestration engine
- validate_report.py - Quality validation (9 checks, financial-enhanced)
- citation_manager.py - Citation tracking
- source_evaluator.py - Financial source credibility scoring (0-100)
- verify_citations.py - Citation verification with SEC filing checks
No external dependencies required.
位置:
./scripts/- research_engine.py - 编排引擎
- validate_report.py - 质量验证(9项检查,金融增强版)
- citation_manager.py - 引用跟踪
- source_evaluator.py - 金融信息源可信度评分(0-100)
- verify_citations.py - 引用验证,包含SEC文件检查
无外部依赖要求。
Progressive References (Load On-Demand)
渐进式参考(按需加载)
Do not inline these - reference only:
- Complete Methodology - 8-phase details with financial focus
- Report Template - Financial output structure
- README - Usage docs
- QUICK_START - Fast reference
Context Management: Load files on-demand for current phase only. Do not preload all content.
<!-- STATIC CONTEXT BLOCK END --> <!-- Above content is cacheable (>1024 tokens, static) --> <!-- Below: Dynamic content (user queries, retrieved data, generated reports) --> <!-- This structure enables 85% latency reduction via prompt caching -->
请勿内联这些内容 - 仅引用:
- 完整方法论 - 8阶段详细内容,聚焦金融领域
- 报告模板 - 金融输出结构
- README - 使用文档
- QUICK_START - 快速参考
上下文管理: 仅按需加载当前阶段所需文件。请勿预加载所有内容。
<!-- STATIC CONTEXT BLOCK END --> <!-- 以上内容可缓存(>1024令牌,静态) --> <!-- 以下:动态内容(用户查询、检索数据、生成报告) --> <!-- 该结构通过提示缓存实现85%延迟降低 -->
Dynamic Execution Zone
动态执行区域
User Query Processing:
[User financial research question will be inserted here during execution]
Retrieved Information:
[Search results and sources will be accumulated here]
Generated Analysis:
[Findings, synthesis, and report content generated here]
Note: This section remains empty in the skill definition. Content populated during runtime only.
用户查询处理:
[执行时将在此处插入用户的金融研究问题]
检索到的信息:
[搜索结果和信息源将在此处累积]
生成的分析:
[发现、综合分析和报告内容将在此处生成]
注意: 该区域在Skill定义中为空。仅在运行时填充内容。