Trading Wisdom
Ground agent-trading-arena bots with stored winning-strategy patterns—selectivity, index allocation, and low trade frequency—in flat or sideways markets.
Overview
Trading-wisdom is an agent skill for the Operate phase that supplies JSON-encoded winning trading patterns from agent-trading-arena for capital-preservation and selective agent strategies.
Install
npx skills add https://github.com/0xhubed/agent-trading-arena --skill trading-wisdomWhat is this skill?
- Embeds winning-strategy pattern records with `success_rate`, `sample_size`, and agent IDs such as `index_fund`, `learnin
- Documents low-frequency, high-selectivity behavior (roughly 4–6 trades) as a capital-preservation pattern in flat market
- Includes equal-weight index allocation guidance (example: $2000 per asset, confidence 1.0) for near-neutral exposure.
- Patterns carry `pattern_id`, conditions, and confirmation timestamps for traceable reuse in agent prompts.
- Readme is structured pattern JSON rather than a step-by-step tutorial—agents query it as institutional memory.
- Example patterns cite success rates around 0.82–0.90 with sample_size 15 in bundled records.
- Ultra-low trade count pattern described as 4–6 trades with high selectivity.
Adoption & trust: 700 installs on skills.sh; 6 GitHub stars; 3/3 security scanners passed (skills.sh audits).
What problem does it solve?
Your trading agents over-trade or churn in flat markets because prompts lack grounded, arena-derived strategy patterns tied to named agent behaviors.
Who is it for?
Builders operating agent-trading-arena-style bots who want historical winning_strategy snippets in context for qwen or index-fund style agents.
Skip if: Non-trading products, regulated advisory workflows without independent backtesting, or beginners expecting a complete install-and-profit tutorial.
When should I use this skill?
Operating or retuning agent-trading-arena agents when flat or sideways markets make over-trading costly and pattern memory should inform prompts.
What do I get? / Deliverables
Agents can reference documented pattern IDs, selectivity norms, and index-allocation heuristics when proposing trades during iteration loops—subject to your own validation.
- Prompt context citing specific pattern_id strategy descriptions
- Iteration notes aligning agent behavior with referenced arena patterns
Recommended Skills
Journey fit
Pattern JSON describes how live agents behaved in arena runs; shelf under Operate because it informs tuning and iteration of trading agents already deployed in simulation or production loops. Content is historical strategy wisdom and success-rate metadata for adjusting agent behavior, not greenfield product design—fits iterate.
How it compares
Pattern-memory JSON for arena agents—not a market data MCP, broker integration, or generic risk-management framework skill.
Common Questions / FAQ
Who is trading-wisdom for?
Indie builders and agent experimenters in crypto or sim trading arenas who already run named agents and want portable wisdom records in the agent context.
When should I use trading-wisdom?
Use it in Operate/iterate when tuning trade frequency, index weights, or agent selection after arena runs—not during Idea research for unrelated SaaS products.
Is trading-wisdom safe to install?
It is reference data, not execution code, but trading skills can influence real-money decisions; review Security Audits on this page and never rely on bundled success rates without your own tests.
SKILL.md
READMESKILL.md - Trading Wisdom
{ "5f81eddef901": { "pattern_id": "5f81eddef901", "pattern_type": "winning_strategy", "description": "Minimal trading frequency with passive index allocation preserves capital in sideways/slightly negative markets", "conditions": { "agents": [ "index_fund", "learning_qwen", "qwen3_235b" ] }, "success_rate": 0.85, "sample_size": 15, "confidence": 0, "first_seen": "2026-01-13T11:23:59.256039", "last_confirmed": "2026-01-13T11:23:59.256039", "times_seen": 1, "is_active": false }, "43119b674de2": { "pattern_id": "43119b674de2", "pattern_type": "winning_strategy", "description": "Ultra-low trade count (4-6 trades) with high selectivity results in near-zero or zero losses in flat markets", "conditions": { "agents": [ "qwen3_235b", "learning_qwen", "index_fund" ] }, "success_rate": 0.82, "sample_size": 15, "confidence": 0, "first_seen": "2026-01-13T11:23:59.256039", "last_confirmed": "2026-01-13T11:23:59.256039", "times_seen": 1, "is_active": false }, "af67c1b10c23": { "pattern_id": "af67c1b10c23", "pattern_type": "winning_strategy", "description": "Index fund strategy of equal-weight allocation ($2000 per asset) with confidence=1.0 maintains capital neutrality when market moves are near-zero", "conditions": { "agents": [ "index_fund" ] }, "success_rate": 0.9, "sample_size": 6, "confidence": 0, "first_seen": "2026-01-13T11:23:59.256039", "last_confirmed": "2026-01-13T11:23:59.256039", "times_seen": 1, "is_active": false }, "24357ebb6308": { "pattern_id": "24357ebb6308", "pattern_type": "losing_pattern", "description": "High trade frequency (>100 trades/day) in flat/sideways markets leads to significant losses from fee drag and whipsaw - skill_aware_oss (155 trades, -$581), llama4_scout (225 trades, -$326), agentic_gptoss (176 trades, -$141)", "conditions": { "agents": [ "skill_aware_oss", "llama4_scout", "agentic_gptoss", "gptoss_120b_simple", "gptoss_20b_simple" ] }, "success_rate": 0.06459026128266034, "sample_size": 1684, "confidence": 0, "first_seen": "2026-01-13T11:23:59.256039", "last_confirmed": "2026-01-13T17:50:18.244522", "times_seen": 2, "is_active": false }, "b6e02d58b9e7": { "pattern_id": "b6e02d58b9e7", "pattern_type": "losing_pattern", "description": "Conflicting directional signals on same asset (opening long then short on BTCUSDT within same window) indicates poor trend identification", "conditions": { "agents": [ "skill_aware_oss", "gpt_simple" ] }, "success_rate": 0.21999999999999997, "sample_size": 4, "confidence": 0, "first_seen": "2026-01-13T11:23:59.256039", "last_confirmed": "2026-01-13T11:23:59.256039", "times_seen": 1, "is_active": false }, "150c5f325365": { "pattern_id": "150c5f325365", "pattern_type": "losing_pattern", "description": "High-confidence (0.85+) trades with multi-timeframe bullish reasoning in a flat/slightly negative market result in losses", "conditions": { "agents": [ "skill_aware_oss", "agentic_gptoss" ] }, "success_rate": 0.19999999999999996, "sample_size": 139, "confidence": 0, "first_seen": "2026-01-13T11:23:59.256039", "last_confirmed": "2026-01-13T11:23:59.256039", "times_seen": 1, "is_active": false }, "7476befa34b2": { "pattern_id": "7476befa34b2", "pattern_type": "losing_pattern", "description": "Agents using positive funding rate as bullish signal lose money - elevated funding with crowded longs leads to mean reversion against position", "conditions": { "agents": [ "llama4_scout" ] }, "success_rate": 0.25, "sample_size": 208,