EVClaw provides an end-to-end live-agent trading framework that lets an OpenClaw agent build cycle contexts, validate proposed trade candidates, and execute orders on Lighter (crypto perps) and Hyperliquid (HIP3 stocks). It supports automated cycle processing, manual trade execution, signal inspection, position management, dynamic risk sizing, and incident triage for agent-driven trading workflows.
Use EVClaw when you need an agent to: run periodic trading cycles and record candidate proposals, execute manual or reviewed trade plans, view actionable signals and positions, or manage trading configuration and risk limits across supported venues. Ideal for teams experimenting with agent-driven execution or building strategy-as-code around cycle files.
cli.py, executor.py, main.py, trading_brain.py).skill.yaml.Best with OpenClaw agent runtimes that support Python-based skills and CLI invocation; integrates with agent workflows that can call live_agent.py and cli.py for execution and analysis.
EVClaw is a live-agent trading skill for crypto perps (Lighter) and HIP3 stocks (Hyperliquid) with cycle-driven signals, risk management, and incident response. Scripts are well-structured with dry-run defaults and argparse, but most require the ai_trader_db module and a populated SQLite DB that aren't bundled. Only 2 of 9 scripts ran successfully in isolation. The SKILL.md is comprehensive with good architecture docs, references, and clear command structure, but the skill demands heavy external setup (env vars, tracker APIs, exchange keys).
ai_trader_dbsqlite3 DB (ai_trader.db)EVClaw runtime environment with tracker APIsNo malicious patterns found. Seed import script downloads from GitHub releases with SHA256 verification and safe tar extraction. Reset script has proper dry-run default and pre-backup. Cron install script auto-registers jobs but is intentional tooling. The skill is domain-specific trading infrastructure — well-engineered but not portable or usable without the full EVClaw runtime stack.