# Alpha Discovery

> Automated alpha mining: Scan market data → Generate factors → Score → Fuse signals.

- Skill: `knuckles-team/alpha-discovery` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add knuckles-team/alpha-discovery`
- Raw SKILL.md: https://api.skillmd.com/api/skills/knuckles-team/alpha-discovery/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Product & Planning
- Author: Knuckles-Team (https://skillmd.com/u/knuckles-team)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/knuckles-team/alpha-discovery

---

# Alpha Discovery Workflow

## Workflow Execution Steps

### Step 1: data-ingest
Fetch market data for target universe.
Tool: `emerald_market_data(action="historical", symbol=..., period="1y")`

### Step 2: feature-engineering
Generate alpha factors with IC/IR scoring.
Tool: `emerald_signals(action="alpha", ticker=...)`

### Step 3: regime-detection
Classify current market regime (Bull/Bear/Sideways/Crisis).
Tool: `emerald_signals(action="regime", ticker=...)`

### Step 4: signal-fusion
Bayesian fusion of all signal sources.
Tool: `emerald_signals(action="fuse", ticker=...)`

### Step 5: kg-persist
Store AlphaFactor and TradingSignal nodes in KG.
Tool: `graph_write(action="add_node", node_type="AlphaFactor", ...)`

## Execution

Run this workflow as a dependency-ordered DAG. Steps with no unmet `depends_on` run in parallel; dependents run after their prerequisites complete.

- **Run first (in parallel):** Step 1 — data-ingest; Step 2 — feature-engineering; Step 3 — regime-detection; Step 4 — signal-fusion; Step 5 — kg-persist

**Execution:** If graph-os is reachable, offload the whole DAG via `graph_orchestrate action=execute_workflow` (or the `kg-delegate` skill) for true parallel/swarm execution. Otherwise execute the steps natively in dependency order: run steps with no unmet `depends_on` in parallel, then their dependents.