Jakeschincariol/trader-skill

What it is

Trader‑skill is a collection of Claude skills (plugins) written in pure Python that let Claude help you build and manage a paper‑trading portfolio. Each skill performs a single, well‑defined task – e.g. copying a hedge‑fund’s 13‑F filing, monitoring news, budgeting positions, checking risk, generating broker tickets, tracking congressional trades, etc. The tools only read public data sources (SEC EDGAR, House Clerk PTR PDFs, news RSS, Yahoo Finance) and never place orders automatically; a human must confirm every trade.

How to get it

1. In Claude – paste the GitHub URL and run install skill.
2. As a Claude Code plugin – use the marketplace commands shown in the README.
3. Manually – clone the repo and copy the trader‑* folders into ~/.claude/skills/ (or into a project’s .claude/skills/).

The code runs on Python 3.8+ with no external dependencies; just set your SEC user‑agent in trader/config.json or via SEC_USER_AGENT.

Core capabilities (the eleven commands)

Command	Purpose	What it does
/trader-copycat	Reverse‑engineer a filer	Downloads a 13‑F‑HR, extracts holdings, computes weight changes, writes targets.json.
/trader-news	News monitor	Pulls headlines and 8‑K filings for every ticker you hold, flags relevant items.
/trader-budget	Budget manager	Allocates a dollar amount to each target, enforces cash buffer and per‑stock caps, outputs orders.json.
/trader-pelosi	Congressional‑trade tracker	Fetches House PTR PDFs as soon as they appear, parses them into rows, alerts on new filings.
/trader-analyst	Earnings data extractor	Retrieves XBRL facts (revenue, EPS, margins, growth) and the latest earnings release for a ticker.
/trader-screener	Custom screener	Applies user‑defined rules (rules.json) to metrics (metrics.json) and ranks passing stocks.
/trader-risk	Risk gate	Validates each proposed order against user limits (size, position, cash, blocked tickers) and returns PASS/BLOCK.
/trader-broker	Order executor (manual)	Formats a broker‑compatible ticket, shows it to the user, and after a “yes” sends it to the configured brokerage connector (Alpaca MCP by default, paper account).
/trader-tax	Tax‑loss harvesting helper	Tracks cost‑basis lots, identifies loss‑harvest candidates, respects wash‑sale windows.
/trader-alerts	Price‑alert service	Fires when a price target, stop, or market open/close occurs, using the NYSE calendar.
/trader-rebalance	Portfolio rebalancer	Detects drift from target weights (beyond a user‑defined band) and proposes corrective trades.

Typical workflow

1. Copy a fund – /trader-copycat → targets.json.
2. Size positions – /trader-budget → orders.json.
3. Risk check – /trader-risk (blocks any violating order).
4. Confirm & send – /trader-broker shows tickets; you type yes; Claude calls your broker.
5. Ongoing – schedule /trader-news, /trader-pelosi, /trader-alerts.
6. Quarterly – repeat copycat and run /trader-rebalance.

Running the code directly

The repo ships small offline test fixtures, so you can try the commands locally, e.g.:

python3 trader-copycat/edgar13f.py find buffett
python3 trader-news/news.py --tickers AAPL GOOGL --hours 24
python3 trader-pelosi/ptr.py index --member pelosi
python3 -m unittest discover -s tests -v

All scripts use only the Python standard library.

Limitations & safety notes

• Not financial advice – the tools only perform data retrieval and arithmetic.
• Paper‑first – default config uses Alpaca’s paper‑trading endpoint; real‑money keys must be supplied via environment variables.
• Data latency – 13‑F filings are up to 45 days old; congressional PTRs are disclosed within 45 days of the trade.
• No automatic execution – orders are sent only after an explicit user confirmation in Claude.
• Public‑source fragility – relies on SEC EDGAR, House Clerk PDFs, RSS feeds, and an unofficial Yahoo Finance endpoint; any change could break a script.

License & authorship

MIT‑licensed, authored by Jake Schincariol (https://opusjake.ai).