stock-research - SKILL.md Agent Skill

name: stock-research description: Interactive deep-dive analysis on any US or Korean stock. Performs multi-signal analysis covering technicals, fundamentals, sector context, and momentum. Generates a probability-weighted forecast with option to log as a formal prediction. Triggers on keywords like analyze [ticker], research [ticker], what about [ticker], [ticker] 분석, 종목 분석, should I buy, tell me about.

Stock Research

Perform a comprehensive multi-signal analysis on a single stock, covering fundamentals, technicals, sector context, and momentum. Generate a probability-weighted directional forecast.

When to Use

When user asks about a specific stock ("What about NVDA?", "삼성전자 어때?")
When user wants a deep-dive before making a trading decision
When user wants to compare a stock's signals before creating a prediction

Prerequisites

bin/stock-cli must be executable (uses uv-managed environment)
No API keys required for Korean data
Optional: FMP_API_KEY for enhanced US fundamentals

Workflow

1. Identify Stock and Market

Parse the user's request to determine:

Ticker: Extract from user input (e.g., "NVDA", "005930", "삼성전자")
Market: Determine US or KR. If a Korean company name is given, use search to find the ticker:

bin/stock-cli search "삼성전자" --market KR

2. Gather Data

Price history (90 days of context):

bin/stock-cli price NVDA --market US --days 90

From the JSON output, calculate:

Current price and short-term trend (5-day, 20-day direction)
52-week high/low proximity (when enough data)
Volume trend (above/below recent average)
Approximate moving averages from close prices (10, 20, 50 day)

Fundamentals:

bin/stock-cli fundamentals NVDA --market US

Context — existing predictions for this ticker:

bin/stock-cli predict list --ticker NVDA --limit 5

Context — broader market for sector signal:

bin/stock-cli price SPY --market US --days 30
# For a semi stock, also fetch semi ETF
bin/stock-cli price SMH --market US --days 30

3. Multi-Signal Analysis

Score each dimension 0-100:

1. Technical Signal (weight: 25%)

Trend: price vs 20-day and 50-day MA
Momentum: accelerating or decelerating moves
Support/resistance: proximity to key levels
Volume confirmation: trend moves on expanding volume

2. Fundamental Signal (weight: 20%)

Valuation: P/E relative to sector (from fundamentals output)
Growth: implied from price trajectory
Quality: margins, balance sheet (when FMP available)

3. Sector Signal (weight: 20%)

Is the sector leading or lagging?
Rotation phase: early/mid/late cycle
Relative strength vs sector peers

4. Momentum Signal (weight: 20%)

Price momentum: 1W, 1M, 3M returns from price data
Volume momentum: increasing/decreasing participation
Breakout/breakdown proximity

5. Sentiment Signal (weight: 15%)

Recent price action patterns
Cross-market correlations
Market-wide risk sentiment (check SPY/KOSPI context)

3b. Bull / Bear / Judge synthesis (before committing a direction)

Don't jump from the signal table straight to a DIRECTION. Run a short adversarial pass so the forecast isn't one-sided:

Thesis (commit first): one line — your directional lean and the single biggest risk to it.
Bull case: the 2-3 strongest specific reasons it goes up (cite the numbers from the signals above).
Bear case: the 2-3 strongest specific reasons it goes down — genuinely adversarial. If you can't write a real one, say so.
Judge: weigh both into the final DIRECTION / CONFIDENCE. Keep the surviving bear point in the KEY RISKS block below. If the bear case beats the thesis, change the direction — don't anchor on the opening lean.

4. Present the Forecast

═══════════════════════════════════════════════
  STOCK RESEARCH: [TICKER] ([NAME])
  Market: [US/KR] | Price: [PRICE] | Date: [DATE]
═══════════════════════════════════════════════

  SIGNAL BREAKDOWN:
  ┌─────────────┬───────────┬──────────┐
  │ Signal      │ Reading   │ Score    │
  ├─────────────┼───────────┼──────────┤
  │ Technical   │ [BULL/..] │ [0-100]  │
  │ Fundamental │ [BULL/..] │ [0-100]  │
  │ Sector      │ [BULL/..] │ [0-100]  │
  │ Momentum    │ [BULL/..] │ [0-100]  │
  │ Sentiment   │ [BULL/..] │ [0-100]  │
  └─────────────┴───────────┴──────────┘

  COMPOSITE SCORE: [0-100]
  DIRECTION: [BULL / BEAR / NEUTRAL]
  CONFIDENCE: [0.50-0.95]
  TIMEFRAME: [1W/2W/1M]

  TARGET: [price] ([+X%])
  STOP: [price] ([-X%])

  REASONING:
  [3-5 sentences synthesizing the signals into a coherent thesis]

  KEY RISKS:
  - [Risk 1]
  - [Risk 2]
═══════════════════════════════════════════════

5. Offer to Log Prediction

After presenting the analysis, ask:

"Want me to log this as a formal prediction? It will be tracked for accuracy scoring."

If yes, invoke:

bin/stock-cli predict create \
  --ticker NVDA \
  --market US \
  --direction BULL \
  --confidence 0.72 \
  --timeframe 1W \
  --entry-price 125.50 \
  --target-price 135.00 \
  --stop-price 120.00 \
  --reasoning "<synthesized thesis>" \
  --signals technical,momentum,sector \
  --source INTERACTIVE

Return the prediction ID to the user so they can reference it later.

Confidence Calculation

Composite score maps to direction:

0-35: BEAR
36-64: NEUTRAL
65-100: BULL

Confidence: 0.50 + (abs(composite - 50) / 50) * 0.45

Composite 50 → confidence 0.50 (coin flip)
Composite 75 → confidence 0.725
Composite 100 → confidence 0.95

Korean Stock Specifics

Default timeframe: 2W (not 1W) due to lower liquidity
Widen stop-loss by ~20% vs US equivalent
Samsung Electronics (005930) and SK Hynix (000660) are proxies for global semi demand
Consider won/dollar exchange rate impact on export-heavy companies
Chaebol discount: Korean conglomerates typically trade at lower P/E than global peers