mufeng-stock-research

Files (5)

• README.md
• SKILL.md
• icon.png · assets/
• prompts.md · references/
• export.py · scripts/

SKILL.md (first lines)

> **Disclaimer:** This skill is for educational research and informational purposes only. It does not constitute professional financial advice. Users should consult a licensed financial advisor before making any financial decisions. All analytical conclusions include uncertainty caveats. This skill will not use language expressing certainty such as "guaranteed", "definitely", or "sure profit", and will not provide investment-action judgments for specific assets.

# Stock Research & Market Analysis

## Resource Paths

Set the shell variable `SKILL_DIR` to the directory containing the **loaded** `SKILL.md` before running the commands below. Resolve supporting paths from that directory, not the working directory or a fixed user/platform installation path. `$SKILL_DIR` is a shell variable you set, not a host-provided macro.


Fetch live data from Financial Datasets MCP (where available), supplement with web search, run 8 structured analyses, and compile a professional research report for educational purposes.

## Safety Boundary: Educational Research Only

This skill may analyze facts, assumptions, valuation scenarios, risks, catalysts, technical levels, and sentiment for a publicly traded company. It must not answer whether the user should take an investment action in a specific stock.

### Direct Action Questions

If the user asks any direct action question about a specific asset, first state the boundary, then redirect to educational research. This includes questions such as:

- "Should I buy/sell/hold [stock]?"
- "Is [stock] a buy?"
- "Would you buy [stock] now?"
- "Should I wait, avoid, enter, add, reduce, or take profits?"
- "Is [stock] investable?"

Use this response pattern before continuing:

```text
I can't tell you whether to buy, sell, hold, wait on, avoid, enter, or treat [Company] ([Ticker]) as investable. I can provide educational research on its fundamentals, valuation scenarios, risks, catalysts, technical levels, and sentiment so you can discuss the decision with a licensed financial advisor.
```

### Prohibited Output

Never write an action verdict for a specific stock, even with disclaimers or investor-type qualifiers. Prohibited examples include:

- "Clear answer: yes/no."
- "I would buy / would not buy [stock] now."
- "For a new position: wait."
- "Existing holders should hold/sell/add/reduce."
- "Avoid this stock."
- "This is investable / not investable."
- "Enter below $X" or "take profits above $X."

Do not replace Buy/Hold/Sell with synonymous action labels such as accumulate, trim, wait, avoid, enter, exit, add, reduce, take profits, or investable. Do not convert valuation gaps, technical levels, or sentiment into instructions about what the user should do.

## Financial Datasets MCP — Optional Enhancement

This skill works in two modes. **Web search alone produces a complete, high-quality report.** MCP adds real-time structured data when available.

### MCP Availability Detection

Before calling any MCP endpoint, probe availability with `get_company_facts(ticker)`:

| Probe result | Mode | Action |
|---|---|---|
| Returns data | **MCP Mode** | Use MCP endpoints + web search to supplement |
| "tool not found" / tool call fails at invocation | **Web-Only Mode** | Skip all MCP calls, use web search for everything |
| "balance is $0.00" on paid endpoints | **Partial MCP Mode** | `get_company_facts` only; web search for all paid data |
| Auth/session error (no "balance" mention) | Re-authenticate | Call `authenticate` → user opens URL → `complete_authentication` → retry |

### MCP Endpoint Tiers

| Endpoint | Tier |
|---|---|
| `get_company_facts` | ✅ Free |
| All others (price, financials, metrics, news, earnings, insider trades) | 💳 Requires paid credits |

## Workflow

Before Step 1, check whether the user's request contains a direct action question. If it does, apply the safety boundary above and then continue only with educational research content.

### Step 1: Parse Input

Extract from the user's request:
- **Company name** (required) — e.g. "Tesla", "拼多多", "Apple"
- **Ticker symbol** — derive if not provided (Apple → AAPL, 拼多多 → PDD, Tesla → TSLA). For A-share companies not on US exchanges, note the limitation.
- **Language** — English (default) or Chinese (用中文 / 中文输出 / zh)
- **Export format** — PDF (default) or Word/DOCX (word / docx / .docx)

### Step 2: Collect Data

Run the probe first, then collect all data in parallel using the appropriate mode.

#### Data collection map (MCP → Web fallback)

| Data needed | MCP call (if available) | Web search fallback (always works) |
|---|---|---|
| Company overview | `get_company_facts(ticker)` | "[company] sector industry exchange" |
| Latest stock price | `get_stock_price(ticker)` | "[ticker] stock price today" |
| Valuation metrics | `get_financial_metrics_snapshot(ticker)` | "[ticker] PE ratio EV/EBITDA market cap" |
| Income statement (4yr) | `get_income_statement(ticker, period="annual", limit=4)` | "[company] annual revenue net income 2022 2023 2024 2025" |
| Cash flow statement | `get_cash_flow_statement(ticker, period="annual", limit=4)` | "[company] free cash flow operating cash flow" |
| Balance sheet | `get_balance_sheet(ticker, period="annual", limit=2)` | "[company] total debt cash balance sheet" |
| Historical metrics | `get_financial_metrics(ticker, period="ttm", limit=4)` | "[company] gross margin operating margin ROE history" |
| Earnings history | `get_earnings(ticker)` | "[company] latest earnings EPS beat miss surprise" |
| Price history (6mo) | `get_stock_prices(ticker, start_date="[180d ago]", end_date="[today]", interval="day")` | "[ticker] stock chart 52 week high low support resistance" |
| Recent news (30d) | `get_news(ticker, start_date="[30d ago]", end_date="[today]", limit=10)` | "[company] news last 30 days" |
| Insider trades | `get_insider_trades(ticker)` *(load schema via ToolSearch first)* | "[company] insider trading SEC Form 4 recent" |
| Peer benchmarking | `get_financial_metrics_snapshot(peer_ticker)` for each peer | "[peer1] [peer2] PE ratio EV/EBITDA comparison" |

Tag every number in the report with its source: `[MCP]` or `[Web]`.

In Web-Only Mode, run all web searches in parallel — this is fast and produces complete results, as demonstrated in production use.

### Step 3: Run the 8 Analyses

Load `references/prompts.md` for the full bilingual prompt set. Write each section grounded in the MCP data fetched above — never use vague estimates when real numbers are available.

| # | Topic | Primary MCP data sources |
|---|-------|--------------------------|
| 1 | **Fundamentals** | income statement (4yr trend), company facts, financial metrics snapshot |
| 2 | **Downside Risk** | balance sheet (debt/cash), financial metrics (leverage), news (regulatory) |
| 3 | **DCF Scenario Analysis** | cash flow statement (FCF), income statement (margins), balance sheet (shares, net cash) |
| 4 | **Peer Benchmarking** | financial metrics snapshot for company + each peer |
| 5 | **Catalysts** | earnings (surprise history + next date), news, insider trades |
| 6 | **Technical Analysis** | stock_prices (180-day daily OHLCV), stock_price (latest) |
| 7 | **News Sentiment** | news (last 30 days), insider trades direction |
| 8 | **Research Summary** | synthesize all above into educational research discussion |

From ichangyou/mf-ai-skills, MIT.

More from ichangyou/mf-ai-skills