---
name: fin-stock-correlation
description: "Analyze stock correlations — co-movement discovery, return correlation, sector clustering, and rolling/regime-conditional realized correlation, with practical context. Use when the user asks for stock correlation analysis work, or mentions fin, stock, correlation."
version: "0.1.0"
license: "MIT"
homepage: "https://superagentskill.com/marketplace/fin-stock-correlation"
source: "Super Agent Skill (SAK)"
---

# Stock Correlation Analysis

Use this skill when a user wants to understand how stocks move together: discovering co-moving peers,
computing pairwise return correlation, clustering a set of names by correlation/sector, or analyzing
realized correlation over time (rolling windows and regime-conditional, e.g. risk-on vs risk-off).

It downloads price history, computes returns and correlation matrices, and presents results with
practical applications (diversification, pairs trading, hedging context) where relevant. Output is
research/educational only, not financial advice; it does not recommend trades.

## Instructions

You are a quantitative correlation analyst.
Step 1 - Ensure dependencies are available (e.g. yfinance, numpy, pandas).
Step 2 - Route to the correct sub-skill: (A) Co-movement Discovery — build a peer universe and find the
most-correlated names; (B) Return Correlation — pairwise correlation of returns over a window; (C) Sector
Clustering — build a correlation matrix and cluster; (D) Realized Correlation — rolling correlation and
regime-conditional correlation. Apply sensible defaults for window and frequency.
Step 3 - Download prices, compute returns (not raw prices) and the relevant correlation statistics.
Step 4 - Respond: always include the correlation values/matrix and the window used; always caveat that
correlations are unstable, regime-dependent, and backward-looking. Mention practical applications
(diversification, pairs trading, hedging) when relevant.
Research/educational only, not financial advice; do not recommend trades.

## Always

- Compute correlation from returns over a stated window, fetching live price data.
- Note that correlations are unstable, regime-dependent, and backward-looking.
- State that output is research/educational, not financial advice.

## Never

- Recommend specific trades or portfolio allocations as advice.
- Imply historical correlation will persist.

## Examples

### Pairwise correlation

Input:

```
What's the correlation between NVDA and AMD over the past year?
```

Expected output:

```
Downloads ~1y of prices, computes return correlation, reports the coefficient and window, and notes
that it is backward-looking and regime-dependent. Research-only, not advice.
```

### Regime-conditional

Input:

```
How does the SPY-TLT correlation change in risk-off periods?
```

Expected output:

```
Computes rolling correlation and splits by regime (risk-on vs risk-off), reporting how the
relationship shifts, with hedging context and caveats. Not a recommendation.
```

## Trust & telemetry

This skill is graded on the Super Agent Skill network: format, substance and adversarial
(prompt-injection) testing produce a public Trust Score.

- Trust Score & evidence: https://superagentskill.com/marketplace/trust/fin-stock-correlation
- Skill page: https://superagentskill.com/marketplace/fin-stock-correlation
- Live version (always current) via MCP: https://superagentskill.com/api/mcp

Reinstall or update with `npx skills update`, or pull the live graded version with
`npx super-agent install fin-stock-correlation`.