Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
This collection bundles ten skills that move from raw market news to structured outputs: fetching news, scoring sentiment with FinBERT, forecasting with the Kronos model, tracking how investment signals evolve, and generating professional reports. It serves traders, portfolio managers, and quantitative researchers who need a single agent workflow to monitor A-Share, HK, and US markets. A built-in skill creator also helps developers package their own domain skills.
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Installs: Global OpenCode installation of the complete Awesome Finance Skills repository bundle; the command copies every folder under `skills/*`, including skills beyond this bounded skillPaths selection if present · OpenCode
Before you start
Git must be available for the documented clone command
OpenCode with its global skills directory at `~/.config/opencode/skills/`
Install the runtime dependencies needed by the skills you use: `requests`, `loguru`, `duckduckgo-search`, `pandas`, `akshare`, `yfinance`, `agno`, `torch`, `transformers`, `sentence-transformers`, `numpy`, and `scikit-learn`; `sqlite3` is documented as built in
Skills using stored data require the bundled database manager and, where stated, a correctly initialized `DatabaseManager`
alphaear-predictor additionally requires the Kronos model and an embedding model; its project-root `exports/models` directory must exist when using trained news-projector weights, and only trusted `kronos_news_*.pt` checkpoints should be placed there
After cloning, run this terminal command from the directory containing the new `Awesome-finance-skills` directory to copy the complete skills bundle into OpenCode's global skills directory.
`{source_id}`: a valid source identifier, such as `cls` or `weibo`
`{count}`: the number of stories to retrieve
Runtime setup: `requests`, `loguru`, and the bundled local database support
Use the alphaear-news skill to fetch the latest hot financial news from source `{source_id}` with a count of `{count}`. Verify that the source ID is valid using the skill's source reference, then return a concise list of headlines and available source information. Do not publish or save anything.
`{company_or_code}`: an A-share, Hong Kong, or US company name or ticker code, such as `Moutai` or `600519`
Runtime setup: the skill's documented Python dependencies and stock-table database support
Use the alphaear-stock skill to fuzzy-search for `{company_or_code}`. Return the matching ticker candidates as code-and-name pairs, and briefly identify which candidate is most likely if the query is ambiguous. Do not retrieve price history yet.
Use the alphaear-sentiment skill's LLM-based analysis mode to analyze this financial text: `{text}`. Return only JSON with `score` from -1.0 to 1.0, `label` as `positive`, `negative`, or `neutral`, and a brief `reason`. Do not update any database.
Claude
Codex
Gemini CLI
OpenClaw
OpenCode
Antigravity