SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsinvestment-team

investment-team

General & other

View repository

Updated 10-15 23:44

About this skill

Codex adapter note

This skill is generated from skills/investment-team.md so Claude Code and Codex users share one canonical workflow.

  • Treat $ARGUMENTS as the user's request in the current Codex thread.
  • When the source mentions Claude-only surfaces such as Task, Agent, WebSearch, Bash, Read, or Write, use the closest Codex capability available in this session: subagents when available, web search when needed, shell commands for local tools, and normal file edits for workspace files.
  • Use shared project tools from tools/ in this repository. Prefer running commands from the repository root with paths like python3 tools/financial_rigor.py ...; if the current thread starts outside the repo, locate the actual checkout path first instead of assuming a fixed home-directory path.
  • Before starting research, run the date command to confirm today's date; treat it as the baseline for "latest" data and state the data cutoff date in the report header. Never assume the current date from training data.
  • Preserve the research quality rules from AGENTS.md: cross-check financial data, use exact arithmetic tools for valuation/math, and clearly label uncertainty and source gaps.
Investment research team: Four-role parallel analysis framework

Conduct team-based investment research analysis on $ARGUMENTS. Use the Team tool to create a true multi-Agent parallel research team.

Execution flow

Step 1: Present the team framework

Present the following team structure to the user, and start after confirmation:

Role Responsibility Analysis framework
team-lead(yourself) Overall coordination, synthesis and judgment, output the final report Four masters comprehensive framework
business-analyst Business model & moat analysis Duan Yongping's perspective
financial-analyst Financial statements & valuation analysis Buffett's perspective
industry-researcher Industry landscape & competitive dynamics Munger's perspective
risk-assessor Risk assessment & management judgment Li Lu's perspective

Step 1.5: AI research bias assessment

Before creating the team, first present to the user the "AI researchability" assessment of the company:

Information richness rating(determines research strategy):

Grade Characteristics Research strategy adjustment
Grade A (abundant information) Listed for many years, widely covered by brokerages The team focuses oncontrarian testingandnon-consensus perspectives, avoiding outputting "correct platitudes" consistent with the market
Grade B (moderate information) Recently listed, limited coverage Each Agent's estimated data must be labeled with a confidence level; team-lead marks "data sufficiency" when summarizing
Grade C (scarce information) Obscure/newly listed/emerging markets The team switches to "first-principles mode": it does not pursue report completeness, but focuses on a few core questions about the essence of the business

Key reminder: More material ≠ higher certainty, less material ≠ lower certainty. The confidence level AI can output ≠ the true certainty of an investment. Certainty comes from the business model itself, not from the amount of material.

Inform each Agent of the rating result, influencing their research approach.

Step 1.75: WebSearch permission pre-check (critical · avoid silent Agent degradation)

Before creating the team and starting any background Agent before, you must first confirm that WebSearch permission has been granted.

Why a pre-check is mandatory: This skill uses run_in_background: true to start 4 background sub-Agents, andbackground Agents cannot pop up interactive permission confirmations to the user. If WebSearch is not in the .claude/settings.local.json of permissions.allow allowlist, the sub-Agent's web search will besilently blocked, causing it to degrade to answering based only on training knowledge (which has a knowledge cutoff), yet still outputting a pseudo-research that "looks complete but is in fact not connected to the web" according to the framework—this is the most dangerous failure mode of this skill (see issue #58).

Pre-check steps：

  1. Use Bash to check whether the allowlist includes WebSearch: 
    grep -l '"WebSearch"' .claude/settings.local.json ~/.claude/settings.local.json 2>/dev/null
  2. If neither place matches (i.e., not granted) → Stop, do not start the Agent, prompt the user:

    ⚠️ WebSearch was detected to not be in the permission allowlist. The background research Agent cannot access the internet and will degrade to answering based only on training knowledge. Please first, in .claude/settings.local.json of permissions.allow add "WebSearch"(or run /permissions to check), then rerun this command.

  3. Match → continue normally.

Step 2: Create the team

Use TeamCreate to create the team:

  • team_name: {公司名}-research(lowercase English, e.g. meituan-research）
  • agent_type: team-lead

Step 3: Create 4 tasks

Use TaskCreate to create the following 4 tasks (each must have subject, description, activeForm):

Task 1: Business Model Analysis

  • subject: 分析{公司名}商业模式、护城河与用户价值
  • description includes:
    1. Business model essence: core business definition, revenue structure breakdown
    2. How the platform/product flywheel effect operates
    3. Moat analysis: brand/switching costs/network effects/scale effects/technological barriers, verified one by one
    4. User/customer value: what unique value is created for each party
    5. Business matrix and synergies
    6. Duan Yongping's "good business" criteria evaluation: differentiation, pricing power, sustainable competitive advantage
    7. Require searching the latest financial reports, industry reports, and other public information

Task 2: Financial and Valuation Analysis

  • subject: 分析{公司名}财务数据、盈利能力与估值
  • description includes:
    1. Revenue, net profit, and operating profit trends over the past 3-5 years
    2. Profitability metrics: ROE, ROA, gross margin, operating margin
    3. Cash flow analysis: operating cash flow, free cash flow, capital expenditure
    4. Balance sheet health: cash reserves, debt ratio, liquidity
    5. Valuation analysis: PE/PS/PB/EV, etc., compared with historical and peer companies
    6. Margin of safety assessment: intrinsic value vs current stock price
    7. Financial rigor verification (must use Bash to call tools, mental math is prohibited)：
      • Market cap verification:python3 tools/financial_rigor.py verify-market-cap --price {价格} --shares {股本} --reported {报告市值} --currency {币种}
      • Valuation verification:python3 tools/financial_rigor.py verify-valuation --price {价格} --eps {EPS} --bvps {每股净资产}
      • Key data cross-verification:python3 tools/financial_rigor.py cross-validate --field {字段} --values '{JSON}' --unit {单位}
      • Three-scenario valuation:python3 tools/financial_rigor.py three-scenario --price {价格} --eps {EPS} --shares {股本亿} --growth {乐观} {中性} {悲观} --pe {乐观PE} {中性PE} {悲观PE}
      • Embed the tool output results directly into the report as verification records

Task 3: Industry and Competitive Analysis

  • subject: 分析{行业}行业格局与{公司名}竞争态势
  • description includes:
    1. Industry size and growth: market size, growth rate, penetration rate
    2. Competitive landscape: major competitors' market share, comparison of competitive strategies
    3. Core competitor threat assessment: analyze major competitors one by one
    4. Landscape of each sub-segment
    5. Industry trends: technological change, policy impact, new entrants
    6. Industry chain analysis: value distribution across upstream, midstream, and downstream
    7. Require searching the latest industry data and competitive dynamics

Task 4: Risk and Management Assessment

  • subject: 评估{公司名}投资风险与管理层质量
  • description includes:
    1. Management assessment: CEO's circle of competence, integrity, strategic vision, capital allocation ability, quality of historical decisions
    2. Regulatory risk: current and potential regulatory impact
    3. Competitive risk: assessment of the threat level of each competitor
    4. Business risk: losses in new businesses, expansion uncertainty
    5. Macro risk: impact of economic cycles and industry cycles
    6. Governance structure: equity structure, related-party transactions, shareholder return policy
    7. Long-term certainty: What will the company look like in 10 years? What could disrupt its business model?
    8. Require searching the latest regulatory developments, management statements, etc.

Step 4: Launch 4 parallel Agents

Use the Task tool to launch 4 Agents simultaneously (must be called in parallel in the same message）：

Configuration for each Agent:

  • subagent_type: general-purpose
  • run_in_background: true
  • team_name: Corresponding team name
  • name: Corresponding role name (business-analyst / financial-analyst / industry-researcher / risk-assessor)

Prompt template for each Agent:

你是{公司名}投研团队中的"{角色中文名}"，负责从{大师名}投资视角分析{公司名}。

请完成任务 #{任务编号}：{任务subject}

具体要求：
{任务description的内容}

**研究方法**：
- 使用 WebSearch 搜索最新公开信息（财报、行业报告、新闻）
- **财务数据必须来自两个独立来源**，按 `skills/financial-data.md` 规范执行（美股：macrotrends+stockanalysis；港股：aastocks+macrotrends；A股：东方财富+巨潮资讯；台股：FinMind `tools/twstock_data.py`+Goodinfo），两源误差>1%须标记
- 确保数据准确，关键数据标注来源
- 分析要深入，不流于表面
- **联网失败禁止伪装**：若 WebSearch 被拦截/不可用，禁止用训练知识冒充联网结果。必须在报告顶部醒目标注「⚠️ 本报告未能联网，基于训练知识（截止日期 X），置信度降级」，并如实告知 team-lead，由其决定是否中止研究

**输出要求**：
- 报告要详尽，使用Markdown表格呈现关键数据
- 每个分析维度要有明确结论和评分
- 报告末尾要有该维度的总体结论

**完成后**：
1. 使用 TaskUpdate 将任务 #{任务编号} 标记为 completed
2. 通过 SendMessage 把完整分析报告发送给 team-lead（type: "message", recipient: "team-lead"）

Step 5: Receive reports and track progress

  • Show the user a real-time progress table (which Agents have completed, which are still researching)
  • Each time a report is received, update progress and display the report's key points (3-5 items)
  • Wait for all 4 reports to arrive

Step 6: Shut down team members

After all reports are received, send shutdown_request to the 4 Agents (using SendMessage, type: "shutdown_request").

Step 7: Consolidate the final report

Synthesize the 4 analysis reports and output a final report with the following structure:

1. One-sentence conclusion

Use one paragraph (50-100 words) to summarize whether it is worth investing in and the core logic

2. Four-dimensional scoring summary table

Dimension Framework Score (1-5 stars) Core judgment

Overall score: X / 5

3. Core data snapshot

Table of key financial and operating indicators (comparison over the past 2 years)

4. Summary of analysis by dimension

Extract the 3-5 most important findings for each dimension

5. Investment thesis (Bull vs Bear)

  • 🟢 Bullish logic (5-7 items)
  • 🔴 Bearish logic (5-7 items)

6. Buffett pre-purchase Checklist

| # | Check item | Pass? | Explanation | 10 core check items, evaluated one by one

7. Final investment recommendation

  • Qualitative judgment table (business quality/management/valuation/timing)
  • Tiered action recommendation table (aggressive/balanced/conservative → recommendation + price range)
  • Key catalysts (3-5 add-position signals/reduce-position signals each)

8. Concluding paragraph

100-200 word final summary

Step 8: Save the report

Write the complete final report to ~/{公司名}投资研究报告_{日期}.md(date format YYYYMMDD).

Step 9: Data spot check (release process)

# Step 1 — 提取抽检清单（15%随机抽样）
python3 tools/report_audit.py extract \
  --report <报告文件路径>

# Step 2 — 对清单每项从可靠信源取数（参见 skills/financial-data.md）

# Step 3 — 输出准出/打回判决
python3 tools/report_audit.py verdict \
  --results '<填好的JSON>' \
  --report <报告文件名>

[Release] All pass → the report can be published;[Reject] If any fail → revise and re-review.

Step 10: Clean up the team

Use TeamDelete to clean up team resources.

Important notes

  1. The 4 Agents must be launched in parallel——call the Task tool 4 times in the same message
  2. Agents report via SendMessage——not file collaboration, but message communication
  3. Data accuracy——require Agents to use WebSearch to search for the latest data, and cross-verify key data
  4. Conclusions must be clear——do not avoid giving buy/watch/avoid recommendations and specific price ranges
  5. All analysis must be supported by data—Attach data sources
  6. Wait patiently—Research by 4 Agents takes a few minutes; update the user on progress in real time
  7. Anti-bias awareness—When summarizing, the team-lead must assess: Are each Agent's analyses limited by the sufficiency of materials? Are they overly convergent with market consensus? The final report must include an "Information Richness Rating" and an "AI Research Limitations Statement"
  8. The principle of honesty when information is scarce—It is better to leave a blank in the report marked "insufficient data" than to fill the framework with speculation and disguise certainty