awesome-finance-skills: Turn Your Agent Into a Market Analyst

Financial questions are where general-purpose agents fall apart fastest: they hallucinate tickers, recite stale prices, and cannot tell you which news actually moved a market. RKiding/Awesome-finance-skills (2.8k stars) is a plug-and-play skill collection — the README's pitch is "transform your AI agent into a Wall Street analyst in seconds" — that gives an agent live news aggregation, real stock data, sentiment scoring, and prediction grounded in actual data sources.

Why This Skill Matters

The pack ships eight alphaear-* skills, each a distinct capability: alphaear-news aggregates hot news from 10+ sources (Cailian, WSJ, Weibo, and Polymarket among them), alphaear-stock covers A-Share, Hong Kong, and US markets with ticker search, OHLCV data, and fundamentals, and alphaear-sentiment scores text with FinBERT or an LLM on a -1.0 to +1.0 scale. The remaining skills round out the pipeline — logic-chain visualization that renders transmission-chain diagrams explaining market impact, a Kronos-based predictor with news-aware adjustments, plus search, reporting, and signal skills.

Two design choices stand out. The README is bilingual (English and Chinese), reflecting the A-Share focus alongside US and HK coverage. And the skills are individually installable — you can wire just the news skill into a general-purpose agent without adopting the whole financial suite.

Installation

Install one skill via the skills CLI:

npx skills add RKiding/Awesome-finance-skills@alphaear-news

Or search the collection interactively:

npx skills find "alphaear"

Manual installation clones the repo and copies the skill folders your agent reads — each folder must contain its SKILL.md:

git clone https://github.com/RKiding/Awesome-finance-skills.git
cp -r Awesome-finance-skills/skills/* ~/.config/opencode/skills/

The README's install-path table covers Claude Code, Codex, OpenCode, and OpenClaw, at both project and global scopes.

Real Workflow: Explain a Market Move

The README's own example is a cross-market question that needs news, data, and reasoning together:

Analyze how the gold crash affects A-shares

The agent combines alphaear-news (recent gold coverage from the aggregated sources), alphaear-stock (current A-Share positions), and alphaear-logic (a transmission-chain diagram of the impact path) — an answer grounded in retrieved data rather than a model's memory of markets.

Real Workflow: Score Sentiment Before a Decision

Point alphaear-sentiment at recent headlines or a watchlist and get scores on the documented -1.0 to +1.0 scale. Combined with alphaear-stock fundamentals, that gives a repeatable pre-analysis routine for a watchlist — ask the agent to score, sort, and explain the outliers.

Tips

• Install per capability: the news skill alone upgrades a general agent; the full suite is for dedicated finance workflows.
• The README links a live demo (DeepEar Lite at deepear.vercel.app) and positions DeepEar as the full automated analysis framework if you outgrow the skills.
• A-Share coverage is a first-class citizen here — unusual among agent skill collections and the reason the Chinese-language docs are not an afterthought.

When Not to Use This

None of this is investment advice or execution infrastructure — there is no brokerage integration, and the prediction skill is a model forecast, not a guarantee. Treat outputs as research inputs. And if your focus is purely US markets with English sources, the A-Share-oriented news aggregation may cover more than you need.

See the leaderboard for more skills.