feat: add quant layer, portfolio-review, and strategy-backtest skills

fundamental-analysis: added statistical risk layer
- normality test (Jarque-Bera) — validates Sharpe/VaR reliability
- unit root test (ADF) — validates technical analysis applicability
- rolling skew/kurtosis — tail risk monitoring
- interpretation rules for crash risk detection

portfolio-review (NEW): portfolio health check and similarity search
- HRP optimization, correlation matrix, risk parity weights
- t-SNE clustering for hidden correlations
- stock similarity search for diversification
- rule-engine BUY_MORE/HOLD/SELL per holding

strategy-backtest (NEW): historical strategy validation
- SMA crossover, RSI mean-reversion, buy-and-hold, momentum
- comparison framework with Sharpe, max DD, win rate
- validation workflow for trade-analyze recommendations

Coverage: 67% → 79% of API endpoints (104/131)

openclaw-skills/fundamental-analysis/SKILL.md

@@ -38,6 +38,18 @@ curl -sk "$BASE/api/v1/stock/{TICKER}/sortino?days=365"
 # Omega ratio (full distribution gain/loss)
 curl -sk "$BASE/api/v1/stock/{TICKER}/omega?days=365"
 
+# Normality test (are returns normally distributed? If not, Sharpe/VaR underestimates risk)
+curl -sk "$BASE/api/v1/stock/{TICKER}/normality?days=365"
+
+# Unit root test (does the price have a trend, or is it random walk?)
+curl -sk "$BASE/api/v1/stock/{TICKER}/unitroot?days=365"
+
+# Rolling skewness (negative skew = increasing downside tail risk)
+curl -sk "$BASE/api/v1/stock/{TICKER}/rolling/skew?days=365&window=30"
+
+# Rolling kurtosis (high = fat tails, extreme moves more likely)
+curl -sk "$BASE/api/v1/stock/{TICKER}/rolling/kurtosis?days=365&window=30"
+
 # Share statistics (float, shares outstanding, short % of float)
 curl -sk "$BASE/api/v1/stock/{TICKER}/share-statistics"
 
@@ -81,10 +93,17 @@ curl -sk "$BASE/api/v1/calendar/earnings?start_date=$(date +%Y-%m-%d)&end_date=$
 - Beta: {value} — [defensive <0.8 / market ~1.0 / aggressive >1.2]
 - CAPM alpha: {%} — [outperforming / underperforming vs market]
 - Sharpe (1Y): {value} — [poor <0.5 / good >1.0 / excellent >2.0]
-- Sortino (1Y): {value} — [context vs Sharpe]
+- Sortino (1Y): {value} — Sortino > Sharpe suggests positive skew (good)
 - Max drawdown (1Y): {%}
 - Omega ratio: {value}
 
+### Statistical Risk (Quant Layer)
+- Normality: {Jarque-Bera p-value} — [normal (p>0.05): Sharpe/VaR reliable / non-normal (p<0.05): use Sortino/Omega instead]
+- Unit root (ADF): {p-value} — [trend present (p<0.05): technicals valid / random walk (p>0.05): technicals unreliable]
+- Rolling skew (30d): {latest value} — [negative = growing downside tail / positive = upside tail / near 0 = symmetric]
+- Rolling kurtosis (30d): {latest value} — [>3 = fat tails, extreme moves likely / <3 = thin tails, well-behaved]
+- ⚠️ If non-normal + negative skew + high kurtosis = **elevated crash risk**, reduce position size
+
 ### Institutional & Ownership
 - Float: {shares} ({% of total})
 - Short % of float: {%} — [low <5% / moderate / high >15%]
@@ -113,4 +132,7 @@ curl -sk "$BASE/api/v1/calendar/earnings?start_date=$(date +%Y-%m-%d)&end_date=$
 - Flag **earnings within 7 days** as a major catalyst/risk
 - Sortino > Sharpe for stocks with asymmetric return profiles
 - High short interest + upcoming catalyst = potential squeeze
-- Keep under 500 words
+- If normality test fails: Sharpe/VaR are unreliable, emphasize Sortino and Omega instead
+- If unit root test fails (random walk): de-emphasize technical signals in your assessment
+- Negative rolling skew + high kurtosis = crash risk flag — recommend smaller position
+- Keep under 600 words

openclaw-skills/portfolio-review/SKILL.md

@@ -0,0 +1,140 @@
+---
+name: portfolio-review
+description: "Portfolio health check — optimization, correlation, risk parity, clustering, similarity analysis. Use when user asks about portfolio allocation, diversification, rebalancing, or finding similar stocks."
+user-invocable: true
+metadata: { "openclaw": { "emoji": "💼", "requires": { "bins": ["curl"] } } }
+---
+
+# Portfolio Review
+
+Professional portfolio analysis. Think like a portfolio manager at a wealth management firm.
+
+**Trigger**: User mentions portfolio, allocation, diversification, rebalancing, "how should I weight", or "find stocks like X".
+
+## Mode A: Portfolio Health Check
+
+When user provides holdings or references MEMORY.md portfolio:
+
+```bash
+BASE=https://invest-api.k8s.home
+
+# Rule-engine analysis (BUY_MORE/HOLD/SELL per holding with confidence)
+curl -sk -X POST "$BASE/api/v1/portfolio/analyze" \
+  -H "Content-Type: application/json" \
+  -d '{"holdings": [{"symbol": "AAPL", "shares": 100, "buy_in_price": 150.0}, ...]}'
+
+# HRP optimal weights (Hierarchical Risk Parity)
+curl -sk -X POST "$BASE/api/v1/portfolio/optimize" \
+  -H "Content-Type: application/json" \
+  -d '{"symbols": ["AAPL", "MSFT", "GOOGL", ...], "days": 365}'
+
+# Correlation matrix (identify hidden correlations)
+curl -sk -X POST "$BASE/api/v1/portfolio/correlation" \
+  -H "Content-Type: application/json" \
+  -d '{"symbols": ["AAPL", "MSFT", "GOOGL", ...], "days": 365}'
+
+# Risk parity weights (equal risk contribution)
+curl -sk -X POST "$BASE/api/v1/portfolio/risk-parity" \
+  -H "Content-Type: application/json" \
+  -d '{"symbols": ["AAPL", "MSFT", "GOOGL", ...], "days": 365}'
+
+# Cluster analysis (which stocks behave similarly)
+curl -sk -X POST "$BASE/api/v1/portfolio/cluster" \
+  -H "Content-Type: application/json" \
+  -d '{"symbols": ["AAPL", "MSFT", "GOOGL", ...], "days": 180}'
+
+# Per-stock risk metrics
+for SYMBOL in AAPL MSFT GOOGL; do
+  curl -sk "$BASE/api/v1/stock/$SYMBOL/performance?days=365"
+done
+```
+
+### Report Structure
+
+```
+## Portfolio Review — {date}
+
+### Holdings Summary
+| Symbol | Shares | Cost | Current | P&L | P&L% | Signal |
+|--------|--------|------|---------|-----|-------|--------|
+| {sym} | {n} | ${cost} | ${current} | ${pnl} | {%} | {BUY_MORE/HOLD/SELL} |
+
+### Total Portfolio
+- Total value: ${sum}
+- Total P&L: ${sum} ({%})
+- Positions: {count}
+
+### Diversification Analysis
+- Correlation highlights:
+  - Highest pair: {A} ↔ {B} = {r} — ⚠️ if >0.8 = redundant exposure
+  - Lowest pair: {A} ↔ {B} = {r} — ✅ good diversification
+- Clusters found: {n} groups
+  - Cluster 1: [{symbols}] — {interpretation, e.g. "tech megacap"}
+  - Cluster 2: [{symbols}] — {interpretation}
+- Diversification score: [well-diversified / concentrated / sector-heavy]
+
+### Optimal Allocation
+| Symbol | Current Weight | HRP Weight | Risk Parity | Action |
+|--------|---------------|------------|-------------|--------|
+| {sym} | {%} | {%} | {%} | [increase/decrease/hold] |
+
+### Risk Metrics per Holding
+| Symbol | Sharpe | Volatility | Max DD | Beta |
+|--------|--------|-----------|--------|------|
+| {sym} | {val} | {%} | {%} | {val} |
+
+### Recommendations
+1. {Top priority action with reasoning}
+2. {Second action}
+3. {Third action}
+
+### ⚠️ Alerts
+- {Any BUY_MORE or SELL signals from rule engine}
+- {Highly correlated pairs that should be reduced}
+- {Overweight positions vs optimal}
+```
+
+## Mode B: Find Similar/Different Stocks
+
+When user asks "find stocks like X" or "what's similar to X":
+
+```bash
+# Find most similar stocks
+curl -sk -X POST "$BASE/api/v1/portfolio/similar" \
+  -H "Content-Type: application/json" \
+  -d '{"symbol": "{TICKER}", "universe": ["AAPL","MSFT","GOOGL","AMZN","META","NVDA","TSLA","JPM","V","WMT","JNJ","PG","XOM","CVX","HD"], "days": 180, "top_n": 5}'
+
+# Cluster the target with potential alternatives
+curl -sk -X POST "$BASE/api/v1/portfolio/cluster" \
+  -H "Content-Type: application/json" \
+  -d '{"symbols": ["{TICKER}","AAPL","MSFT","GOOGL","AMZN","META","NVDA","TSLA","JPM","V","WMT","JNJ"], "days": 180}'
+```
+
+### Report Structure
+
+```
+## Stocks Similar to {TICKER}
+
+### Most Similar (by return correlation)
+| Rank | Symbol | Correlation | Why Similar |
+|------|--------|-------------|-------------|
+| 1 | {sym} | {r} | {interpretation} |
+
+### Most Different (diversification candidates)
+| Rank | Symbol | Correlation | Why Different |
+|------|--------|-------------|---------------|
+| 1 | {sym} | {r} | {interpretation} |
+
+### Recommendation
+- To add exposure like {TICKER}: consider {top similar}
+- To hedge {TICKER}: consider {most different}
+```
+
+## Rules
+
+- Always read MEMORY.md for existing holdings before analysis
+- Correlation > 0.8 = redundant — flag for reduction
+- Correlation < 0.2 = good diversification pair
+- HRP is generally better than equal-weight for risk-adjusted returns
+- Risk parity = each position contributes equal risk (good for conservative portfolios)
+- Keep under 500 words

openclaw-skills/strategy-backtest/SKILL.md

@@ -0,0 +1,130 @@
+---
+name: strategy-backtest
+description: "Strategy backtesting — test SMA crossover, RSI, buy-and-hold, momentum strategies against historical data. Use when user wants to validate a trading idea or compare strategies."
+user-invocable: true
+metadata: { "openclaw": { "emoji": "🧪", "requires": { "bins": ["curl"] } } }
+---
+
+# Strategy Backtest
+
+Professional backtesting for strategy validation. Think like a quant researcher — data over intuition.
+
+**Trigger**: User says "backtest", "test this strategy", "would this have worked", "compare strategies", or wants to validate a trade-analyze recommendation.
+
+## Available Strategies
+
+### 1. SMA Crossover (trend-following)
+
+When short SMA crosses above long SMA → buy. Crosses below → sell.
+
+```bash
+BASE=https://invest-api.k8s.home
+
+curl -sk -X POST "$BASE/api/v1/backtest/sma-crossover" \
+  -H "Content-Type: application/json" \
+  -d '{"symbol": "{TICKER}", "short_window": 20, "long_window": 50, "days": 365, "initial_capital": 10000}'
+```
+
+Best for: trending markets, medium-term holds.
+Weak in: sideways/choppy markets (many false signals).
+
+### 2. RSI Mean Reversion
+
+Buy when RSI < oversold threshold, sell when RSI > overbought threshold.
+
+```bash
+curl -sk -X POST "$BASE/api/v1/backtest/rsi" \
+  -H "Content-Type: application/json" \
+  -d '{"symbol": "{TICKER}", "period": 14, "oversold": 30, "overbought": 70, "days": 365, "initial_capital": 10000}'
+```
+
+Best for: range-bound stocks, mean-reverting behavior.
+Weak in: strong trends (catches falling knives).
+
+### 3. Buy and Hold (benchmark)
+
+Always run this as the baseline comparison.
+
+```bash
+curl -sk -X POST "$BASE/api/v1/backtest/buy-and-hold" \
+  -H "Content-Type: application/json" \
+  -d '{"symbol": "{TICKER}", "days": 365, "initial_capital": 10000}'
+```
+
+### 4. Momentum (multi-stock rotation)
+
+Rank stocks by recent performance, hold top N, rebalance periodically.
+
+```bash
+curl -sk -X POST "$BASE/api/v1/backtest/momentum" \
+  -H "Content-Type: application/json" \
+  -d '{"symbols": ["AAPL","MSFT","GOOGL","AMZN","NVDA","META","TSLA","JPM","V","WMT"], "lookback": 60, "top_n": 3, "rebalance_days": 30, "days": 365, "initial_capital": 10000}'
+```
+
+Best for: diversified portfolios, capturing sector rotation.
+
+## Standard Workflow
+
+**Always run all 3 single-stock strategies + buy-and-hold for comparison:**
+
+```bash
+# Run all 4 in one go
+curl -sk -X POST "$BASE/api/v1/backtest/buy-and-hold" -H "Content-Type: application/json" -d '{"symbol":"{TICKER}","days":365,"initial_capital":10000}'
+curl -sk -X POST "$BASE/api/v1/backtest/sma-crossover" -H "Content-Type: application/json" -d '{"symbol":"{TICKER}","short_window":20,"long_window":50,"days":365,"initial_capital":10000}'
+curl -sk -X POST "$BASE/api/v1/backtest/rsi" -H "Content-Type: application/json" -d '{"symbol":"{TICKER}","period":14,"oversold":30,"overbought":70,"days":365,"initial_capital":10000}'
+```
+
+## Report Structure
+
+```
+## {TICKER} Strategy Backtest — {date}
+### Period: {start_date} to {end_date} ({days} days)
+### Initial Capital: $10,000
+
+### Strategy Comparison
+| Strategy | Return | Sharpe | Max DD | Win Rate | Trades |
+|----------|--------|--------|--------|----------|--------|
+| Buy & Hold | {%} | {val} | {%} | N/A | 1 |
+| SMA 20/50 | {%} | {val} | {%} | {%} | {n} |
+| RSI 14/30/70 | {%} | {val} | {%} | {%} | {n} |
+
+### Winner: {strategy name}
+- Outperformed buy-and-hold by: {%}
+- Key advantage: {why it worked for this stock}
+
+### Equity Curve Summary
+- Buy & Hold final: ${value}
+- Best strategy final: ${value}
+- Worst drawdown period: {date range}
+
+### Strategy Suitability for {TICKER}
+- Stock behavior: [trending / mean-reverting / choppy]
+- Best fit: {strategy} because {reason}
+- Avoid: {strategy} because {reason}
+
+### ⚠️ Backtest Caveats
+- No transaction costs or slippage included
+- Past performance ≠ future results
+- Optimized parameters may overfit
+- Consider out-of-sample testing (different time period)
+```
+
+## Validation Workflow (after /trade-analyze)
+
+When used to validate a trade-analyze recommendation:
+
+1. Run buy-and-hold for baseline
+2. If trade-analyze recommended BUY based on technical signals:
+   - Run SMA crossover to see if trend-following would have worked
+   - Run RSI to see if mean-reversion entries would have worked
+3. Compare Sharpe ratios and max drawdowns
+4. Conclusion: "The data {supports / does not support} the trade-analyze recommendation because {reason}"
+
+## Rules
+
+- **Always include buy-and-hold as benchmark** — any strategy must beat it
+- Sharpe > 1.0 = good risk-adjusted returns
+- Max drawdown > 20% = strategy needs tighter risk management
+- Win rate < 40% can still be profitable if average win >> average loss
+- If all strategies underperform buy-and-hold → the stock rewards patience, not trading
+- Keep under 500 words