invest-committee (Investment Committee)

Turn “analyze TSMC” into a source-cited investment-committee memo: fundamentals, institutional consensus, technical levels, risk flags, and a position-sizing action plan.

An Agent Skill for Claude Code that helps retail investors research specific stocks with more discipline. It covers Taiwan / US / Hong Kong / China A-share markets and routes data sources by ticker. It supports research only — it never places orders.

Language note: reports are currently generated in Traditional Chinese (zh-TW). The workflow and methodology are market-agnostic.

Three things you can rely on

Every figure needs a source. Prices, financials, and flow data are cited; unavailable values are labeled as missing.
Data gaps are visible. Stale data, source fallbacks, and insufficient candle history are disclosed instead of presented as a complete conclusion.
Research never becomes execution. The skill does not call trading APIs; its output is research support, not investment advice.

Try this first (one stage)

In Claude Code, say:

Analyze NVDA technicals

This runs only the relevant technical-analysis stage, so it is a faster and lower-token way to see whether the workflow fits you. When you are ready, try a full deep dive with “Analyze NVDA in depth.”

What are you trying to do?

Your goal	Say this in Claude Code
Research one stock from fundamentals through risk	“幫我分析台積電” / “Analyze NVDA in depth”
Check only technicals or risk	“看 2330 技術面” / “掃 NVDA 風險”
Check the health of a watchlist	“健檢自選股”
Compare two candidate stocks	“比較台積電和聯電” / “Compare TSMC vs UMC”
Build a Taiwan day-trading playbook	“力成當沖” / “6239 當沖劇本”

A full analysis delivers a source-cited research report and an HTML dashboard with position sizing, entry/exit levels, stop-loss, and a risk checklist. Day-trading support is Taiwan-only.

If auto-triggering misses, say “use invest-committee to analyze …” explicitly. Theme- or industry-level stock screening is outside this skill’s scope; build a concrete ticker list first, then research each name here.

The six-stage workflow

Data collection → fundamentals & growth quality → institutional consensus → money flow & technical timing → risk radar → action plan

Technical indicators are calculated by scripts/indicators.py, never estimated mentally. When a primary source is unavailable, the workflow falls back through available sources and records why in the report.

Fastest start

git clone https://github.com/HenryLinyy/invest-committee.git
cd invest-committee
chmod +x install.sh
./install.sh

The installer adds the skill, wires up the Taiwan data source (twstock MCP), and creates an investor profile, asking for confirmation at every step.

Requirements: Claude Code and Python 3. Taiwan market data requires Docker. Real-time US/HK/China-A data can optionally use Futu OpenD; without it, the skill falls back to free sources.

For manual and Windows installation, data-source details, and token controls, see INSTALL.en.md.

Disclaimer

For research and educational purposes only — not investment advice. Nothing in the output constitutes an offer to buy or sell any security. Trading decisions and outcomes are your own responsibility.

Invest Committee

README