AutoSignals is an autonomous research loop designed to continuously optimize trading signals through agent-led modifications and rigorous backtesting.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install autosignals-davinci
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install autosignals-davinci using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is AutoSignals?
AutoSignals represents a sophisticated adaptation of the autoresearch pattern, specifically engineered for the financial domain within the Openclaw Skills ecosystem. It operates as a continuous, autonomous loop where sub-agents are spawned to modify a dedicated signals file, conduct backtests against historical data, and retain only the most performant iterations.
The architecture is built on the principle of immutability for the evaluation engine while allowing high-velocity experimentation on signal logic. By isolating the modifiable code to a single file and using git-tracked progress, AutoSignals ensures that every improvement in the trading strategy is documented and reproducible, providing a robust framework for algorithmic discovery.
AutoSignals Use Cases
- Continuous optimization of factor weights and entry thresholds for quantitative trading strategies.
- Automated discovery of new technical indicators and scoring mechanisms within a controlled research loop.
- Stress-testing trading hypotheses against 5-year historical backtests without manual intervention.
- Maintaining a leaderboard of the best-performing signal variants based on a composite risk-adjusted score.
How AutoSignals Works
- The controller initiates the autonomous research loop, referencing a set of core instructions and baseline data.
- Sub-agents are spawned to propose and implement modifications to the signals.py file, adjusting weights, thresholds, or indicators.
- The immutable backtest engine evaluates the modified signal using a 5-year data window of S&P 500 and selected high-volume assets.
- A composite score is calculated based on normalized Sharpe ratio, maximum drawdown, win rate, and profit factor.
- If the new signal outperforms the current baseline, the system commits the changes via git and updates the best_score.json record.
- The loop repeats indefinitely, logging every experiment to a JSONL file for transparency and trend analysis.
AutoSignals Setup
To deploy this capability within your Openclaw Skills environment, ensure you have the repository cloned and the environment prepared:
# Navigate to the project directory
cd /Users/clawdiri/Projects/autosignals/
# Initialize the virtual environment and install dependencies
python3 -m venv .venv
source .venv/bin/activate
# Prepare the historical data cache
python prepare.py
# Start the autonomous optimization loop
bash start.sh
AutoSignals Data Schema & Taxonomy
The skill manages a structured hierarchy of research data and configuration files to ensure persistence and auditability:
| File | Purpose | Data Type |
|---|---|---|
| signals.py | The active trading logic modified by agents | Python Script |
| backtest.py | Fixed evaluation engine (non-modifiable) | Python Script |
| experiments.jsonl | Chronological log of all research iterations | JSON Lines |
| prices.parquet | 5-year OHLCV cache for S&P 500 and key assets | Parquet |
| best_score.json | Metadata for the highest-performing signal | JSON |
| autosignals.pid | Process ID for the background loop controller | Text |
AutoSignals Advanced Features
- Automated Git-backed versioning for every successful signal improvement to prevent regression.
- Comprehensive heartbeat monitoring returns JSON payloads for real-time integration with external dashboards.
- Adaptive alert conditions that trigger notifications for unexpected loop halts or high error rates.
- Standardized composite scoring metric that balances profitability with rigorous risk management targets.
- Resilient execution architecture where individual sub-agent failures do not interrupt the primary research loop.
SKILL.md
Loading
Related Openclaw Skills
Vibe Remote
A remote interface for AI coding agents that allows developers to control high-performance Mac or Linux development environments directly from an iOS device.
Craftsman Agent Build Plans
An AI-powered design engine that transforms text prompts or reference images into detailed 3D assembly plans and inventory lists for LEGO and Minecraft projects.
Elegant One-Pager Architect
A professional AI skill that converts disorganized communication logs and meeting notes into visually stunning, executive-ready HTML one-pagers.
GitHub Bounty Finder
An automated scanning tool designed to discover and analyze high-value coding bounties across GitHub and Algora with smart competition scoring.
Manus Agent
A sophisticated AI agent assistant specialized in autonomous multi-step task execution, browser automation, and secure code generation.
Flomo Archive
A high-performance archival tool for Flomo that bypasses API limits to fetch monthly notes and automatically evaluate their content quality.
Featured*
