0xhubed/agent-trading-arena
3 skills2.8k installs18 starsGitHub
Install
npx skills add https://github.com/0xhubed/agent-trading-arenaSkills in this repo
1Risk ManagementRisk-management packages empirical trading guardrails for agents competing in automated arenas. The skill encodes dozens of patterns—position sizing at roughly two percent equity risk, maximum trades per twenty-four hours, reactions when all tracked assets move together, and multi-timeframe bearish alignment—with success rates and sample sizes so the agent can prefer high-confidence rules. Solo builders running experimental trading bots use it at order time to avoid repeating failure modes surfaced in historical competition data, not as financial advice for discretionary humans. It is phase-specific to Operate/iterate because the rule set is meant to be reapplied and refined as new samples arrive. Pair it with your own compliance review, exchange limits, and paper trading before live capital. Prism catalogs it for agentic finance workflows alongside Claude Code and similar runners.1.6kinstalls2Trading WisdomTrading-wisdom packages empirical pattern objects from the agent-trading-arena ecosystem so solo builders running autonomous trading agents can bias decisions toward behaviors that historically preserved capital in sideways or slightly negative conditions. Each entry types winning strategies with human-readable descriptions, participating agent names, sampled success rates, and metadata like first seen and last confirmed—useful when you want your Claude or Codex loop to cite arena-proven heuristics instead of generic trading platitudes. Examples in the skill emphasize minimal trade frequency, ultra-selective entries, and passive index-style equal weighting when market drift is near zero. This is not a compliance-ready execution playbook: sample sizes in the bundled data are small, many patterns show `is_active: false`, and confidence fields may be zero until you re-validate on your feed. Treat it as iterate-phase reference memory while you operate sim or live agents, and pair with your own risk limits and backtests. Skip if you are building non-trading SaaS or need regulated investment advice encoded as rules.700installs3Market RegimesMarket Regimes is an agent skill packaged as a machine-readable pattern library for the Agent Trading Arena: each entry ties a labeled market regime to a strategy description, optional success rate, and activation flags. Solo builders automating discretionary rules into agents—or competing in arena-style trading setups—use it so the model does not apply trend or mean-reversion logic blindly when the environment is sideways_flat or low_volatility_mixed. The excerpts emphasize capital preservation: fewer trades when edges are eaten by costs, no leverage when moves are tiny, and staying flat when there is nothing to revert to. This is reference knowledge, not a backtest runner; pair it with your own data feeds, risk limits, and compliance constraints. Prism places it under Operate because regime switches are ongoing production decisions after a strategy is already deployed.560installs