// the find
RKiding/Awesome-finance-skills
A collection of Awesome Finance Agent Skills for free and easy to start | 一系列开源免费的金融分析Agent Skills
A collection of plug-in skills for AI coding agents (Claude Code, OpenCode, etc.) that give them financial data capabilities: real-time news aggregation, stock data, sentiment scoring, time-series forecasting, and report generation. Primarily oriented toward Chinese markets (A-shares, Cailian, Weibo sources) with some US/HK coverage. Aimed at developers building finance-focused agent workflows.
The skill-per-directory structure is clean and the SKILL.md convention makes it straightforward to drop individual capabilities into any compatible agent without pulling in the whole repo. The Kronos predictor ships an actual trained .pt model file rather than just pointing you at a hosted API, so you can run inference locally. Multi-framework support is genuinely useful — the same skill folder works in Claude Code, OpenCode, and Antigravity without modification. Sentiment scoring via FinBERT with a numeric output range is more actionable than a label-only classifier.
Massive code duplication across skills: `database_manager.py`, `llm/factory.py`, `llm/router.py`, and the full Kronos model implementation are copy-pasted verbatim into at least five skill directories, which means bug fixes don't propagate and the repo is bloated. The 'real-time' news sources are heavily weighted toward Chinese-language outlets (Cailian, Weibo, 财联社) with WSJ as a thin US proxy — someone expecting solid US/EU coverage will be disappointed. No dependency manifest per skill (no `requirements.txt` or `pyproject.toml` in each skill folder), so you won't know what's missing until something ImportErrors at runtime. The Kronos model is presented as a forecasting tool with no backtesting results, confidence intervals, or baseline comparison — treating it as predictive signal in production would be a mistake without independent validation.