{
  "schemaVersion": "1.0",
  "item": {
    "slug": "strategy-workflow",
    "name": "Strategy Workflow",
    "source": "tencent",
    "type": "skill",
    "category": "效率提升",
    "sourceUrl": "https://clawhub.ai/ahuserious/strategy-workflow",
    "canonicalUrl": "https://clawhub.ai/ahuserious/strategy-workflow",
    "targetPlatform": "OpenClaw"
  },
  "install": {
    "downloadMode": "redirect",
    "downloadUrl": "/downloads/strategy-workflow",
    "sourceDownloadUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=strategy-workflow",
    "sourcePlatform": "tencent",
    "targetPlatform": "OpenClaw",
    "installMethod": "Manual import",
    "extraction": "Extract archive",
    "prerequisites": [
      "OpenClaw"
    ],
    "packageFormat": "ZIP package",
    "includedAssets": [
      "SKILL.md",
      "backtest-optimize.md",
      "references/strategy_generation.md"
    ],
    "primaryDoc": "SKILL.md",
    "quickSetup": [
      "Download the package from Yavira.",
      "Extract the archive and review SKILL.md first.",
      "Import or place the package into your OpenClaw setup."
    ],
    "agentAssist": {
      "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.",
      "steps": [
        "Download the package from Yavira.",
        "Extract it into a folder your agent can access.",
        "Paste one of the prompts below and point your agent at the extracted folder."
      ],
      "prompts": [
        {
          "label": "New install",
          "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete."
        },
        {
          "label": "Upgrade existing",
          "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run."
        }
      ]
    },
    "sourceHealth": {
      "source": "tencent",
      "status": "healthy",
      "reason": "direct_download_ok",
      "recommendedAction": "download",
      "checkedAt": "2026-04-30T16:55:25.780Z",
      "expiresAt": "2026-05-07T16:55:25.780Z",
      "httpStatus": 200,
      "finalUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network",
      "contentType": "application/zip",
      "probeMethod": "head",
      "details": {
        "probeUrl": "https://wry-manatee-359.convex.site/api/v1/download?slug=network",
        "contentDisposition": "attachment; filename=\"network-1.0.0.zip\"",
        "redirectLocation": null,
        "bodySnippet": null
      },
      "scope": "source",
      "summary": "Source download looks usable.",
      "detail": "Yavira can redirect you to the upstream package for this source.",
      "primaryActionLabel": "Download for OpenClaw",
      "primaryActionHref": "/downloads/strategy-workflow"
    },
    "validation": {
      "installChecklist": [
        "Use the Yavira download entry.",
        "Review SKILL.md after the package is downloaded.",
        "Confirm the extracted package contains the expected setup assets."
      ],
      "postInstallChecks": [
        "Confirm the extracted package includes the expected docs or setup files.",
        "Validate the skill or prompts are available in your target agent workspace.",
        "Capture any manual follow-up steps the agent could not complete."
      ]
    },
    "downloadPageUrl": "https://openagent3.xyz/downloads/strategy-workflow",
    "agentPageUrl": "https://openagent3.xyz/skills/strategy-workflow/agent",
    "manifestUrl": "https://openagent3.xyz/skills/strategy-workflow/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/strategy-workflow/agent.md"
  },
  "agentAssist": {
    "summary": "Hand the extracted package to your coding agent with a concrete install brief instead of figuring it out manually.",
    "steps": [
      "Download the package from Yavira.",
      "Extract it into a folder your agent can access.",
      "Paste one of the prompts below and point your agent at the extracted folder."
    ],
    "prompts": [
      {
        "label": "New install",
        "body": "I downloaded a skill package from Yavira. Read SKILL.md from the extracted folder and install it by following the included instructions. Tell me what you changed and call out any manual steps you could not complete."
      },
      {
        "label": "Upgrade existing",
        "body": "I downloaded an updated skill package from Yavira. Read SKILL.md from the extracted folder, compare it with my current installation, and upgrade it while preserving any custom configuration unless the package docs explicitly say otherwise. Summarize what changed and any follow-up checks I should run."
      }
    ]
  },
  "documentation": {
    "source": "clawhub",
    "primaryDoc": "SKILL.md",
    "sections": [
      {
        "title": "Strategy Workflow",
        "body": "Comprehensive strategy development workflow for quantitative trading, from hypothesis to validated production deployment."
      },
      {
        "title": "Overview",
        "body": "This skill provides a complete framework for developing, testing, and validating trading strategies. It supports:\n\nHypothesis-driven strategy development\nMulti-GPU backtesting on Vast.ai\nBayesian hyperparameter optimization with Optuna\nWalk-forward validation and out-of-sample testing\nAutomated tearsheet generation"
      },
      {
        "title": "Control Plane (Swarm Orchestration)",
        "body": "Always-on watchdog loops that manage hardware utilization and self-healing:\n\nbash scripts/start_swarm_watchdogs.sh\n\nFor local environments, set explicit paths:\n\nVENV_PATH=/path/to/.venv/bin/activate \\\nRESULTS_ROOT=/path/to/backtests \\\nSTATE_ROOT=/path/to/backtests/state \\\nLOGS_ROOT=/path/to/backtests/logs \\\nbash scripts/start_swarm_watchdogs.sh"
      },
      {
        "title": "Work Plane (Parallel Execution)",
        "body": "Unified wrapper that starts control plane and launches parallel work:\n\nscripts/backtest-optimize --parallel\n\nMulti-GPU, multi-symbol execution:\n\ncd WORKFLOW && ./launch_parallel.sh"
      },
      {
        "title": "Single-Symbol Pipeline",
        "body": "For focused optimization on a single asset:\n\nscripts/backtest-optimize --single --symbol SYMBOL --engine native --prescreen 50000 --paths 1000 --by-regime"
      },
      {
        "title": "1. Hypothesis Formulation",
        "body": "Define your strategy hypothesis in measurable terms:\n\nWhat market inefficiency are you exploiting?\nWhat is the expected holding period?\nWhat are the entry/exit conditions?\nWhat is the target risk-adjusted return?"
      },
      {
        "title": "2. Feature Selection",
        "body": "Identify relevant features for signal generation:\n\nPrice-based (OHLCV, returns, volatility)\nTechnical indicators (EMA, RSI, Bollinger Bands)\nMulti-timeframe features (MTF resampling)\nVolume analysis (PVSRA, VWAP)\nMarket microstructure (order flow, spread)"
      },
      {
        "title": "3. Signal Generation",
        "body": "Convert features into actionable signals:\n\nDirectional bias (trend following, mean reversion)\nEntry conditions (threshold crossings, pattern recognition)\nExit conditions (take-profit, stop-loss, trailing stops)\nPosition sizing rules"
      },
      {
        "title": "4. Position Sizing",
        "body": "Implement risk-aware position sizing:\n\nFixed fractional\nKelly criterion\nVolatility-adjusted\nRegime-dependent scaling"
      },
      {
        "title": "Pre-Flight Validation",
        "body": "MANDATORY before every optimization run:\n\npython validation.py --check-all --data-path DATA_PATH --symbol SYMBOL\n\nValidation checks:\n\nData >= 90 days with no gaps/NaN\nMin trades >= 30 for statistical significance\nMTF resampling implemented correctly\nNo look-ahead bias"
      },
      {
        "title": "Multi-GPU Execution on Vast.ai",
        "body": "Deploy to cloud GPU instances for large-scale parameter sweeps:\n\n# Copy workflow files\nscp -P PORT workflow_files root@HOST:/root/WORKFLOW/\n\n# Run optimization\nssh -p PORT root@HOST \"cd /root/WORKFLOW && python optimize_strategy.py \\\n  --data-path /root/data --symbol SYMBOL --mode aggressive \\\n  --prescreen 5000 --paths 200 --engine gpu\""
      },
      {
        "title": "Prescreening with Vectorized Backtests",
        "body": "Phase 0: GPU-accelerated parameter screening:\n\nGenerate N random parameter combinations\nBatch evaluate on GPU\nFilter by minimum trades (30+)\nReturn top K by Sharpe ratio\n\nPerformance baseline (RTX 5090, 730d lookback, 250k combos): ~4s per mode."
      },
      {
        "title": "Full Backtests with NautilusTrader",
        "body": "Phase 1: Event-driven backtesting for top candidates:\n\nHigh-fidelity simulation with realistic execution\nSlippage and commission modeling\nMulti-asset portfolio backtests"
      },
      {
        "title": "Optuna for Hyperparameter Search",
        "body": "Phase 2: Bayesian optimization with warm-start from prescreening:\n\nimport optuna\n\nstudy = optuna.create_study(\n    direction=\"maximize\",\n    sampler=optuna.samplers.TPESampler(seed=42),\n    pruner=optuna.pruners.MedianPruner()\n)\n\nstudy.optimize(objective, n_trials=1000)"
      },
      {
        "title": "Grid Search vs Bayesian Optimization",
        "body": "MethodUse CaseGrid SearchSmall parameter space, exhaustive coverage neededRandom SearchLarge space, quick explorationBayesian (TPE)Efficient optimization, exploitation/exploration balanceCMA-ESContinuous parameters, smooth objective"
      },
      {
        "title": "Pruning Strategies",
        "body": "MedianPruner: Prune if worse than median of completed trials\nPercentilePruner: Prune bottom X% of trials\nHyperbandPruner: Multi-fidelity optimization\nSuccessiveHalvingPruner: Aggressive early stopping"
      },
      {
        "title": "Distributed Optimization",
        "body": "For large-scale runs, use persistent storage:\n\n# JournalStorage for multi-process\nstorage = optuna.storages.JournalStorage(\n    optuna.storages.JournalFileStorage(\"journal.log\")\n)\n\n# RDBStorage for distributed clusters\nstorage = optuna.storages.RDBStorage(\"postgresql://...\")"
      },
      {
        "title": "Rolling Window Validation",
        "body": "Slide the training/test window through time:\n\n[Train 1][Test 1]\n    [Train 2][Test 2]\n        [Train 3][Test 3]\n\nParameters:\n\ntrain_window: Training period length\ntest_window: Out-of-sample test length\nstep_size: Window advancement increment"
      },
      {
        "title": "Anchored Walk-Forward",
        "body": "Expand training window while sliding test window:\n\n[Train 1      ][Test 1]\n[Train 1 + 2      ][Test 2]\n[Train 1 + 2 + 3      ][Test 3]\n\nUse when historical regime diversity improves model robustness."
      },
      {
        "title": "Epoch Selection Criteria",
        "body": "Intelligent selection of training periods:\n\nRegime-aware: Match training regimes to expected deployment conditions\nVolatility-adjusted: Include both high and low volatility periods\nEvent-inclusive: Ensure major market events are represented\nRecency-weighted: Emphasize recent data while maintaining diversity"
      },
      {
        "title": "Out-of-Sample Testing",
        "body": "Final validation phase:\n\nHold out 20-30% of data for final OOS test\nNo parameter tuning on OOS data\nMonte Carlo stress testing\nRegime-conditional performance analysis"
      },
      {
        "title": "Utilization Targets",
        "body": "CPU utilization target: >= 70%\nGPU utilization target: >= 70%\nNo silent GPU fallback for GPU sweeps"
      },
      {
        "title": "Hardware Watchdog Hooks",
        "body": "Enforced by:\n\nhooks/hardware_capacity_watchdog.py\nscripts/process_auditor.py"
      },
      {
        "title": "Capacity Monitoring",
        "body": "Control plane loops monitor:\n\nWorker health and liveness\nProgress artifact freshness\nResource utilization\nJob queue depth\n\nSelf-healing actions:\n\nAutomatic worker restart on crash\nFill lanes for underutilized resources\nCooldown guardrails to prevent thrashing"
      },
      {
        "title": "Tearsheet Generation",
        "body": "Generate QuantStats-style performance reports:\n\nscripts/generate-tearsheet STRATEGY_NAME \\\n  --trades /path/to/trades.csv \\\n  --capital 10000 \\\n  --output ./tearsheets\n\nSee tearsheet-generator skill for detailed visualization options."
      },
      {
        "title": "PAL MCP Integration",
        "body": "Attach PAL as an MCP server for research/consensus across multiple model providers:\n\nConfig template: config/mcp/pal.mcp.json.example\nDocs: docs/reference/PAL_MCP_INTEGRATION.md\nProviders: OpenRouter, OpenAI, Anthropic, xAI, local models"
      },
      {
        "title": "Documentation",
        "body": "VectorBT Documentation\nNautilusTrader Docs\nOptuna Documentation\nQuantStats"
      },
      {
        "title": "Project References",
        "body": "config/workflow_defaults.yaml - Default configuration\nconfig/model_policy.yaml - Model policy (advisory)\ndocs/guides/SWARM_OPTIMIZATION_RUNBOOK.md - Detailed runbook\nhooks/pipeline-hooks.md - Hook contracts\ndocs/reference/VECTORBT_GRAPH_INGEST.md - VectorBT PRO integration"
      },
      {
        "title": "Results Structure",
        "body": "Backtests/optimizations/{SYMBOL}/{MODE}/\n  best_sharpe/\n    config.json      # Best Sharpe configuration\n    metrics.json     # Performance metrics\n  best_returns/\n  lowest_drawdown/\n  best_winrate/\n  all_trials.json    # All Optuna trials\n  phase0_top500.json # Prescreening results"
      }
    ],
    "body": "Strategy Workflow\n\nComprehensive strategy development workflow for quantitative trading, from hypothesis to validated production deployment.\n\nOverview\n\nThis skill provides a complete framework for developing, testing, and validating trading strategies. It supports:\n\nHypothesis-driven strategy development\nMulti-GPU backtesting on Vast.ai\nBayesian hyperparameter optimization with Optuna\nWalk-forward validation and out-of-sample testing\nAutomated tearsheet generation\nEntry Points\nControl Plane (Swarm Orchestration)\n\nAlways-on watchdog loops that manage hardware utilization and self-healing:\n\nbash scripts/start_swarm_watchdogs.sh\n\n\nFor local environments, set explicit paths:\n\nVENV_PATH=/path/to/.venv/bin/activate \\\nRESULTS_ROOT=/path/to/backtests \\\nSTATE_ROOT=/path/to/backtests/state \\\nLOGS_ROOT=/path/to/backtests/logs \\\nbash scripts/start_swarm_watchdogs.sh\n\nWork Plane (Parallel Execution)\n\nUnified wrapper that starts control plane and launches parallel work:\n\nscripts/backtest-optimize --parallel\n\n\nMulti-GPU, multi-symbol execution:\n\ncd WORKFLOW && ./launch_parallel.sh\n\nSingle-Symbol Pipeline\n\nFor focused optimization on a single asset:\n\nscripts/backtest-optimize --single --symbol SYMBOL --engine native --prescreen 50000 --paths 1000 --by-regime\n\nStrategy Development\n1. Hypothesis Formulation\n\nDefine your strategy hypothesis in measurable terms:\n\nWhat market inefficiency are you exploiting?\nWhat is the expected holding period?\nWhat are the entry/exit conditions?\nWhat is the target risk-adjusted return?\n2. Feature Selection\n\nIdentify relevant features for signal generation:\n\nPrice-based (OHLCV, returns, volatility)\nTechnical indicators (EMA, RSI, Bollinger Bands)\nMulti-timeframe features (MTF resampling)\nVolume analysis (PVSRA, VWAP)\nMarket microstructure (order flow, spread)\n3. Signal Generation\n\nConvert features into actionable signals:\n\nDirectional bias (trend following, mean reversion)\nEntry conditions (threshold crossings, pattern recognition)\nExit conditions (take-profit, stop-loss, trailing stops)\nPosition sizing rules\n4. Position Sizing\n\nImplement risk-aware position sizing:\n\nFixed fractional\nKelly criterion\nVolatility-adjusted\nRegime-dependent scaling\nBacktesting\nPre-Flight Validation\n\nMANDATORY before every optimization run:\n\npython validation.py --check-all --data-path DATA_PATH --symbol SYMBOL\n\n\nValidation checks:\n\nData >= 90 days with no gaps/NaN\nMin trades >= 30 for statistical significance\nMTF resampling implemented correctly\nNo look-ahead bias\nMulti-GPU Execution on Vast.ai\n\nDeploy to cloud GPU instances for large-scale parameter sweeps:\n\n# Copy workflow files\nscp -P PORT workflow_files root@HOST:/root/WORKFLOW/\n\n# Run optimization\nssh -p PORT root@HOST \"cd /root/WORKFLOW && python optimize_strategy.py \\\n  --data-path /root/data --symbol SYMBOL --mode aggressive \\\n  --prescreen 5000 --paths 200 --engine gpu\"\n\nPrescreening with Vectorized Backtests\n\nPhase 0: GPU-accelerated parameter screening:\n\nGenerate N random parameter combinations\nBatch evaluate on GPU\nFilter by minimum trades (30+)\nReturn top K by Sharpe ratio\n\nPerformance baseline (RTX 5090, 730d lookback, 250k combos): ~4s per mode.\n\nFull Backtests with NautilusTrader\n\nPhase 1: Event-driven backtesting for top candidates:\n\nHigh-fidelity simulation with realistic execution\nSlippage and commission modeling\nMulti-asset portfolio backtests\nParameter Optimization\nOptuna for Hyperparameter Search\n\nPhase 2: Bayesian optimization with warm-start from prescreening:\n\nimport optuna\n\nstudy = optuna.create_study(\n    direction=\"maximize\",\n    sampler=optuna.samplers.TPESampler(seed=42),\n    pruner=optuna.pruners.MedianPruner()\n)\n\nstudy.optimize(objective, n_trials=1000)\n\nGrid Search vs Bayesian Optimization\nMethod\tUse Case\nGrid Search\tSmall parameter space, exhaustive coverage needed\nRandom Search\tLarge space, quick exploration\nBayesian (TPE)\tEfficient optimization, exploitation/exploration balance\nCMA-ES\tContinuous parameters, smooth objective\nPruning Strategies\nMedianPruner: Prune if worse than median of completed trials\nPercentilePruner: Prune bottom X% of trials\nHyperbandPruner: Multi-fidelity optimization\nSuccessiveHalvingPruner: Aggressive early stopping\nDistributed Optimization\n\nFor large-scale runs, use persistent storage:\n\n# JournalStorage for multi-process\nstorage = optuna.storages.JournalStorage(\n    optuna.storages.JournalFileStorage(\"journal.log\")\n)\n\n# RDBStorage for distributed clusters\nstorage = optuna.storages.RDBStorage(\"postgresql://...\")\n\nWalk-Forward Validation\nRolling Window Validation\n\nSlide the training/test window through time:\n\n[Train 1][Test 1]\n    [Train 2][Test 2]\n        [Train 3][Test 3]\n\n\nParameters:\n\ntrain_window: Training period length\ntest_window: Out-of-sample test length\nstep_size: Window advancement increment\nAnchored Walk-Forward\n\nExpand training window while sliding test window:\n\n[Train 1      ][Test 1]\n[Train 1 + 2      ][Test 2]\n[Train 1 + 2 + 3      ][Test 3]\n\n\nUse when historical regime diversity improves model robustness.\n\nEpoch Selection Criteria\n\nIntelligent selection of training periods:\n\nRegime-aware: Match training regimes to expected deployment conditions\nVolatility-adjusted: Include both high and low volatility periods\nEvent-inclusive: Ensure major market events are represented\nRecency-weighted: Emphasize recent data while maintaining diversity\nOut-of-Sample Testing\n\nFinal validation phase:\n\nHold out 20-30% of data for final OOS test\nNo parameter tuning on OOS data\nMonte Carlo stress testing\nRegime-conditional performance analysis\nSLOs and Guardrails\nUtilization Targets\nCPU utilization target: >= 70%\nGPU utilization target: >= 70%\nNo silent GPU fallback for GPU sweeps\nHardware Watchdog Hooks\n\nEnforced by:\n\nhooks/hardware_capacity_watchdog.py\nscripts/process_auditor.py\nCapacity Monitoring\n\nControl plane loops monitor:\n\nWorker health and liveness\nProgress artifact freshness\nResource utilization\nJob queue depth\n\nSelf-healing actions:\n\nAutomatic worker restart on crash\nFill lanes for underutilized resources\nCooldown guardrails to prevent thrashing\nTearsheet Generation\n\nGenerate QuantStats-style performance reports:\n\nscripts/generate-tearsheet STRATEGY_NAME \\\n  --trades /path/to/trades.csv \\\n  --capital 10000 \\\n  --output ./tearsheets\n\n\nSee tearsheet-generator skill for detailed visualization options.\n\nMulti-Provider Orchestration\nPAL MCP Integration\n\nAttach PAL as an MCP server for research/consensus across multiple model providers:\n\nConfig template: config/mcp/pal.mcp.json.example\nDocs: docs/reference/PAL_MCP_INTEGRATION.md\nProviders: OpenRouter, OpenAI, Anthropic, xAI, local models\nResources\nDocumentation\nVectorBT Documentation\nNautilusTrader Docs\nOptuna Documentation\nQuantStats\nProject References\nconfig/workflow_defaults.yaml - Default configuration\nconfig/model_policy.yaml - Model policy (advisory)\ndocs/guides/SWARM_OPTIMIZATION_RUNBOOK.md - Detailed runbook\nhooks/pipeline-hooks.md - Hook contracts\ndocs/reference/VECTORBT_GRAPH_INGEST.md - VectorBT PRO integration\nResults Structure\nBacktests/optimizations/{SYMBOL}/{MODE}/\n  best_sharpe/\n    config.json      # Best Sharpe configuration\n    metrics.json     # Performance metrics\n  best_returns/\n  lowest_drawdown/\n  best_winrate/\n  all_trials.json    # All Optuna trials\n  phase0_top500.json # Prescreening results"
  },
  "trust": {
    "sourceLabel": "tencent",
    "provenanceUrl": "https://clawhub.ai/ahuserious/strategy-workflow",
    "publisherUrl": "https://clawhub.ai/ahuserious/strategy-workflow",
    "owner": "ahuserious",
    "version": "0.1.0",
    "license": null,
    "verificationStatus": "Indexed source record"
  },
  "links": {
    "detailUrl": "https://openagent3.xyz/skills/strategy-workflow",
    "downloadUrl": "https://openagent3.xyz/downloads/strategy-workflow",
    "agentUrl": "https://openagent3.xyz/skills/strategy-workflow/agent",
    "manifestUrl": "https://openagent3.xyz/skills/strategy-workflow/agent.json",
    "briefUrl": "https://openagent3.xyz/skills/strategy-workflow/agent.md"
  }
}