How do I install the Ml Classification Project skill?

Run npx skillmds@latest add knuckles-team/ml-classification-project in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Ml Classification Project safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Ml Classification Project?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Ml Classification Project free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Ml Classification Project?

Knuckles-Team (@knuckles-team) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Ml Classification Project by knuckles-team

Name: Ml Classification Project
Author: Knuckles-Team

Ml Classification Project

Build the ML-classification portfolio project: engineer technical-indicator features, train an XGBoost model to predict whether the S&P 500 will be up or down tomorrow, benchmark it against a coin flip, and publish a GitHub repo.

Knuckles-Team 70a4605 2 files · 3.1 KB Updated Sep 17, 2026

File contents

ML Classification Project Workflow

Project 5 of the quant portfolio. Predict next-day direction (not price) and benchmark honestly against a coin flip. See quant-career-docs/reference/ml-for-finance.md.

Steps

Step 1: engineer-features

Agent: quant_researcher Tools: data-science-mcp

Pull S&P 500 history via yfinance and engineer technical-indicator features (returns, moving averages, RSI, volatility, momentum). Ensure features use only past data — no lookahead. Expected: feature-matrix

Step 2: train-xgboost [depends_on: Step 1]

Agent: quant_researcher

Train an XGBoost/LightGBM classifier to predict up/down for the next day. Use time-aware train/test splitting (no shuffling across time). Expected: trained-classifier

Step 3: benchmark-coinflip [depends_on: Step 2]

Agent: quant_researcher

Evaluate out-of-sample accuracy/AUC and benchmark against a 50/50 coin flip and a naive majority-class baseline. Expected: benchmark-comparison

Step 4: evaluate [depends_on: Step 3]

Agent: risk_analyst

If the classifier drives a strategy, report Sharpe, max drawdown, CAGR vs benchmark. State whether the edge survives costs — honest analysis. Expected: evaluation-metrics

Step 5: github-publish [depends_on: Step 4]

Agent: quant_developer

Publish a GitHub repo with README, metrics, the coin-flip benchmark, clean code, and an honest "what didn't work" section. Expected: github-repo-url

Step 6: kg-persist [depends_on: Step 5]

Agent: quant_researcher Tools: graph_write

Persist the project and metrics as typed nodes linked to the portfolio.

Output

An XGBoost direction classifier benchmarked against a coin flip
A published GitHub portfolio repo

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 — engineer-features
After level 0: Step 2 — train-xgboost
After level 1: Step 3 — benchmark-coinflip
After level 2: Step 4 — evaluate
After level 3: Step 5 — github-publish
After level 4: Step 6 — 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.

---name: ml-classification-projectdescription: Build the ML-classification portfolio project: engineer technical-indicator features, train an XGBoost model to predict whether the S&P 500 will be up or down tomorrow, benchmark it against a coin flip, and publish a GitHub repo.---# ML Classification Project WorkflowProject 5 of the quant portfolio. Predict next-day direction (not price) andbenchmark honestly against a coin flip. See`quant-career-docs/reference/ml-for-finance.md`.## Steps### Step 1: engineer-features**Agent**: `quant_researcher`**Tools**: `data-science-mcp`Pull S&P 500 history via yfinance and engineer technical-indicator features(returns, moving averages, RSI, volatility, momentum). Ensure features use onlypast data — no lookahead.Expected: `feature-matrix`### Step 2: train-xgboost [depends_on: Step 1]**Agent**: `quant_researcher`Train an XGBoost/LightGBM classifier to predict up/down for the next day. Usetime-aware train/test splitting (no shuffling across time).Expected: `trained-classifier`### Step 3: benchmark-coinflip [depends_on: Step 2]**Agent**: `quant_researcher`Evaluate out-of-sample accuracy/AUC and benchmark against a 50/50 coin flip and anaive majority-class baseline.Expected: `benchmark-comparison`### Step 4: evaluate [depends_on: Step 3]**Agent**: `risk_analyst`If the classifier drives a strategy, report Sharpe, max drawdown, CAGR vsbenchmark. State whether the edge survives costs — honest analysis.Expected: `evaluation-metrics`### Step 5: github-publish [depends_on: Step 4]**Agent**: `quant_developer`Publish a GitHub repo with README, metrics, the coin-flip benchmark, clean code,and an honest "what didn't work" section.Expected: `github-repo-url`### Step 6: kg-persist [depends_on: Step 5]**Agent**: `quant_researcher`**Tools**: `graph_write`Persist the project and metrics as typed nodes linked to the portfolio.## Output- An XGBoost direction classifier benchmarked against a coin flip- A published GitHub portfolio repo## ExecutionRun 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 — engineer-features- **After level 0:** Step 2 — train-xgboost- **After level 1:** Step 3 — benchmark-coinflip- **After level 2:** Step 4 — evaluate- **After level 3:** Step 5 — github-publish- **After level 4:** Step 6 — 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.

Frequently asked questions