Skip to content
Get packs

Event-Driven Trading Signals

Build sentiment-scored news and event signals and combine them with technical signals for backtests

Source Download skill
Repository
HKUDS/Vibe-Trading
Stars
37,433
Licence
MIT
Files
2 files

What it does

This skill teaches an AI agent how to run an event-driven trading strategy. The agent fetches news and announcements from URLs, scores each item's market sentiment from -1.0 (very bearish) to 1.0 (very bullish) using a fixed scoring prompt, and saves the results in a standard CSV format. It then applies exponential time decay so older events matter less, and blends the event signal with a technical signal (60% technical, 40% event by default) to produce a combined trading signal.

Great for

  • Quantitative researchers testing whether news sentiment adds value to a price-based strategy
  • Developers building backtests that need point-in-time-safe event data
  • Students learning how event signals, decay and look-ahead bias work in trading systems

Skip it if

  • Making live trading decisions: the scores come from an LLM reading news, which can be wrong or biased, and the skill is framed as backtesting support, not investment advice
  • Users without historical news data and a backtest setup: the method relies on correctly dated events, and the skill does not provide a data source

Try asking

  • “Fetch recent earnings announcements for my watchlist, score each one from -1 to 1, and write an event CSV”
  • “Add an event-driven signal with 10-day decay to my existing technical strategy and backtest the 60/40 blend”
  • “Explain why my backtest may have look-ahead bias when using announcements released after market close”
Step-by-step guide: how to use Event-Driven Trading SignalsInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 2 original files, the MIT licence and an attribution note.

    event-driven.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip event-driven.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip event-driven.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Check that each event's date is when it became public, not when it occurred, and use the next trading day for after-close releases
  • Deduplicate events that appear in several news sources so one story does not count multiple times
  • Treat LLM scores as rough estimates and validate the strategy out of sample before relying on it

What's inside

  • example_signal_engine.py
  • SKILL.md

Redistributed unmodified from HKUDS/Vibe-Trading under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.