The Problem
Financial LLM agents often lack ready‑made, domain‑specific capabilities such as live news ingestion, stock‑price retrieval, sentiment scoring, or visual explanations of market logic. Building each of these pieces from scratch consumes time and introduces inconsistency across projects.
What This Does
The repository ships a curated set of “skills”—self‑contained Python packages that augment any LLM‑backed agent with finance‑focused functions. Each skill lives under skills/<skill‑name>/ and follows a uniform layout:
SKILL.md – markdown description and usage notes. scripts/ – the implementation (e.g., newstools.py, stocktools.py, visualizer.py). tests/ – unit tests exercising the public API (e.g., skills/alphaear‑news/tests/testnews.py).
Key examples:
alphaear‑news – aggregates real‑time headlines from ten sources (skills/alphaear-news/scripts/newstools.py). alphaear‑stock – provides ticker lookup, OHLCV, and fundamentals (skills/alphaear-stock/scripts/stocktools.py). alphaear‑predictor – wraps the Kronos time‑series model with news‑aware adjustments (skills/alphaear-predictor/scripts/kronospredictor.py). alphaear‑logic‑visualizer – renders transmission‑chain diagrams as Draw.io XML (skills/alphaear-logic-visualizer/scripts/visualizer.py).
All skills expose a small public API that can be called directly from an agent’s prompt or imported as a module.
How To Use It
Setup
Clone the repo git clone https://github.com/RKiding/Awesome-finance-skills.git cd Awesome-finance-skills
(Optional) Install any listed requirements – none are bundled, so install common deps manually: pip install pandas requests python-dotenv tqdm For the Kronos model you will also need torch and transformers: pip install torch transformers
Installation into an Agent
The README recommends the community‑wide npx skills helper:
Install a single skill, e.g. the news skill
npx skills add RKiding/Awesome-finance-skills@alphaear-news
If the helper is unavailable, copy the desired skill folder into the agent’s skill directory (paths are documented in the README under Integration Guide):
Example for OpenCode
cp -r skills/alphaear-news ~/.config/opencode/skills/
Each skill must contain its SKILL.md; the agent framework will discover it automatically.
Running a Skill
From Python you can import the implementation directly:
from alphaearnews.scripts.newstools import fetchlatest headlines = fetchlatest() print(headlines[:5])
For a full‑pipeline report, combine multiple skills:
from alphaearstock.scripts.stocktools import getprice from alphaearsentiment.scripts.sentimenttools import analyzesentiment from alphaearreporter.scripts.reportagent import generatereport
price = getprice('AAPL') sent = analyzesentiment('Apple earnings beat expectations') report = generatereport(price=price, sentiment=sent) print(report)
All test suites can be executed with pytest (the repo already includes tests/ and skill‑specific tests).
Real‑World Use
A trading desk wants an LLM that can answer “What is the impact of today’s gold price drop on Chinese A‑shares?” The workflow:
from alphaearnews.scripts.newstools import fetchlatest from alphaearsentiment.scripts.sentimenttools import analyzesentiment from alphaearstock.scripts.stocktools import getprice from alphaearreporter.scripts.reportagent import generatereport
news = fetchlatest(topic='gold') sentiment = analyzesentiment(news) price = getprice('600519.SS') # Kweichow Moutai report = generatereport( title='Gold Crash Impact on A‑shares', newssummary=news[:3], sentimentscore=sentiment, stockprice=price, ) print(report)
The LLM can then embed the generated report into its response, delivering a data‑backed answer without custom code per request.
Code Health & Issues
Medium – Missing CI/CD – No .github/workflows, Makefile, or other automation; test execution relies on manual pytest. Low – No explicit dependency file – No requirements.txt or pyproject.toml; users must infer required packages. Low – Large model artifact in repo – kronosnewsv120260101_0015.pt inflates repo size and may cause accidental distribution of binary blobs. Low – Duplicate utility modules – Several skills contain near‑identical utils/llm/ implementations, increasing maintenance overhead. Low – Minimal documentation beyond README – Each SKILL.md provides a brief overview, but function‑level docstrings are sparse, making IDE‑assistance limited. None – License present – LICENSE file included, so reuse is permitted. None – Test coverage – 12 test files cover most entry points; however, integration tests across multiple skills are absent.
The Bottom Line
The repo delivers a practical, plug‑and‑play collection of finance‑focused LLM skills with clear folder conventions and unit tests. It is well‑suited for teams that need to augment existing agents quickly, provided they can manage the manual dependency installation and accept the current lack of CI automation. Organizations looking for a turnkey, production‑grade pipeline should add automated testing, a consolidated dependency manifest, and consider extracting shared utilities to reduce duplication.