メニュー
Planモードの設計・作成・設定を自動化するスキル。AI開発ワークフローのPlanモードを効率的に構築し、チェックポイント管理・進捗監視・エラー回復機能を統合。SO8Tプロジェクト専用に最適化。
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
| name | plan-mode-creation |
| description | Planモードの設計・作成・設定を自動化するスキル。AI開発ワークフローのPlanモードを効率的に構築し、チェックポイント管理・進捗監視・エラー回復機能を統合。SO8Tプロジェクト専用に最適化。 |
| metadata | {"short-description":"Planモード作成を自動化","version":"1.0.0","author":"SO8T Assistant","capabilities":["plan_mode_design","workflow_automation","checkpoint_management","progress_tracking","error_recovery"]} |
SO8Tプロジェクト専用に設計されたPlanモードの作成・設定・管理を自動化する高度なスキル。AIモデル開発の複雑なワークフローを効率的にPlan化し、信頼性の高い実行環境を構築します。
from skills.plan_mode_creation import PlanModeCreator
# AEGISモデル開発Planの自動作成
creator = PlanModeCreator()
plan_config = {
"project_name": "AEGIS-Phi3.5mini-jp Enhancement",
"task_type": "model_training_optimization",
"complexity": "ultra_high",
"resources": {
"gpu_required": True,
"gpu_memory_gb": 12,
"estimated_duration_hours": 48
},
"success_criteria": [
"performance_improvement > 5%",
"statistical_significance < 0.05",
"quantization_loss < 10%"
]
}
# Planモードの自動生成と設定
plan = creator.create_comprehensive_plan(plan_config)
print(f"Plan created: {plan.name}")
print(f"Estimated phases: {len(plan.phases)}")
print(f"Total checkpoints: {plan.total_checkpoints}")
# GGUF量子化最適化Plan
quantization_plan = creator.create_quantization_plan({
"input_model": "AEGIS-Phi3.5mini-jp",
"target_formats": ["Q8_0", "Q4_K_M", "Q3_K_L"],
"calibration_dataset": "so8t_benchmark_data",
"optimization_goals": {
"accuracy_retention": "maximize",
"inference_speed": "maximize",
"model_size": "minimize"
},
"benchmark_validation": True
})
# 最適化されたPlan実行
result = quantization_plan.execute_with_optimization()
# 研究論文自動生成Plan
paper_plan = creator.create_research_paper_plan({
"topic": "SO(8) NKAT Theory and Quadruple Inference",
"target_journal": "arXiv",
"sections": [
"abstract", "introduction", "theoretical_background",
"methodology", "experimental_results", "discussion", "conclusion"
],
"figures": ["theory_diagram", "benchmark_charts", "performance_graphs"],
"citations": ["geometric_dl_papers", "llm_research", "so8t_publications"],
"language": "bilingual" # 英語/日本語
})
# 完全自動論文生成
generated_paper = paper_plan.execute_research_workflow()
# タスクの自動分析
analyzer = creator.analyze_task_requirements(user_input)
print(f"Task complexity: {analyzer.complexity}")
print(f"Required phases: {analyzer.phase_count}")
print(f"Resource requirements: {analyzer.resources}")
# 最適なPlan構造の自動生成
structure = creator.design_plan_structure(analyzer)
print(f"Generated phases: {structure.phases}")
print(f"Checkpoint strategy: {structure.checkpoint_strategy}")
print(f"Error recovery plan: {structure.recovery_plan}")
# Planの詳細設定
config = creator.configure_plan_settings(structure, {
"checkpoint_interval": 180, # 3分
"notification_enabled": True,
"progress_reporting": True,
"error_recovery": True,
"parallel_execution": True
})
# Planの検証と最適化
validator = creator.validate_plan_configuration(config)
optimizer = creator.optimize_plan_performance(validator)
# 最終Planの生成
final_plan = creator.generate_final_plan(optimizer)
# 事前定義済みテンプレートを使用
templates = creator.list_available_templates()
print("Available templates:")
for template in templates:
print(f"- {template.name}: {template.description}")
# テンプレートからPlan作成
plan = creator.create_from_template("aegis_model_training", custom_config)
# カスタムPlanのステップバイステップ構築
builder = creator.create_plan_builder()
builder.add_phase("data_preparation", {
"name": "Data Preparation",
"duration": "2h",
"resources": ["CPU", "Storage"],
"tasks": ["download_dataset", "preprocess_data", "validate_quality"]
})
builder.add_phase("model_training", {
"name": "Model Training",
"duration": "24h",
"resources": ["GPU"],
"checkpoint_interval": 180,
"validation_frequency": 3600
})
builder.add_condition("training_success", "validation_accuracy > 0.85")
builder.add_notification("phase_complete", "email_admin")
plan = builder.build()
# 利用可能なリソースに基づく自動最適化
optimizer = creator.create_resource_optimizer()
optimized_config = optimizer.optimize_for_environment({
"available_gpus": 2,
"gpu_memory_per_gpu": 24, # GB
"cpu_cores": 16,
"available_ram": 64, # GB
"storage_space": 500 # GB
})
print(f"Recommended parallel tasks: {optimized_config.parallel_tasks}")
print(f"Optimal batch size: {optimized_config.batch_size}")
print(f"Memory allocation: {optimized_config.memory_allocation}")
# Planの包括的検証
verifier = creator.create_plan_verifier()
validation_results = verifier.validate_plan(plan, {
"syntax_check": True,
"logic_validation": True,
"resource_consistency": True,
"performance_estimation": True,
"failure_scenario_analysis": True
})
if validation_results.is_valid:
print("Plan validation passed!")
print(f"Estimated success rate: {validation_results.success_probability}%")
else:
print("Validation issues found:")
for issue in validation_results.issues:
print(f"- {issue.severity}: {issue.description}")
# 実行時間の予測と最適化
predictor = creator.create_performance_predictor()
prediction = predictor.predict_execution_time(plan, {
"hardware_profile": "rtx3080_12gb",
"dataset_size": "50gb",
"model_complexity": "high",
"parallel_processing": True
})
print(f"Estimated total time: {prediction.total_hours} hours")
print(f"Estimated cost: ${prediction.compute_cost}")
print(f"Bottleneck analysis: {prediction.bottlenecks}")
# Enhanced Moonshot Pipelineとの統合
moonshot_integration = creator.integrate_with_moonshot({
"pipeline_path": "enhanced_moonshot_pipeline_power_recovery.py",
"integration_points": ["phase_10.7", "phase_10.8", "phase_10.9"],
"shared_resources": True,
"checkpoint_synchronization": True
})
integrated_plan = moonshot_integration.create_integrated_plan()
# GitHub Actionsとの統合
ci_integration = creator.create_ci_integration({
"platform": "github_actions",
"workflow_name": "so8t_model_training",
"trigger_events": ["push", "schedule"],
"environment_variables": {
"MODEL_NAME": "AEGIS-Phi3.5mini-jp",
"QUANTIZATION_LEVELS": "Q8_0,Q4_K_M"
}
})
ci_integration.generate_workflow_file()
plan_mode_creation:
default_template: "so8t_model_development"
checkpoint_interval: 180
max_parallel_tasks: 4
notification_channels: ["email", "slack"]
error_recovery: true
progress_reporting: true
resource_monitoring: true
performance_optimization: true
# SO8Tプロジェクトの最適化設定
so8t_config = {
"statistical_validation": True,
"benchmark_automation": True,
"quantization_optimization": True,
"llm_judge_integration": True,
"paper_generation": True,
"bilingual_support": True
}
creator.apply_project_settings(so8t_config)
# 新しいPlanテンプレートの作成
template_builder = creator.create_template_builder()
template_builder.set_name("custom_so8t_workflow")
template_builder.set_description("SO8T専用カスタムワークフロー")
template_builder.add_phase_template("custom_phase", {
"required_fields": ["task_name", "duration", "resources"],
"optional_fields": ["checkpoint_interval", "validation_criteria"],
"default_values": {
"checkpoint_interval": 300,
"validation_criteria": "accuracy > 0.8"
}
})
custom_template = template_builder.build()
creator.register_template(custom_template)
# カスタムプラグインの開発
class CustomBenchmarkPlugin(PlanPlugin):
def execute(self, context):
# カスタムベンチマーク実行ロジック
results = self.run_custom_benchmarks(context.model_path)
return self.format_results(results)
# プラグイン登録
creator.register_plugin("custom_benchmark", CustomBenchmarkPlugin())
# 詳細デバッグ有効化
creator.enable_debug_mode()
creator.set_log_level("DEBUG")
# Plan作成のステップバイステップ追跡
debug_plan = creator.create_plan_with_debug(user_config)
実装完了: 2026-01-17 01:30:00 機能: Planモード作成スキル実装 ワークツリー名: plan_mode_creation
実装内容:
技術仕様:
使用方法:
from skills.plan_mode_creation import PlanModeCreatorcreator = PlanModeCreator()plan = creator.create_comprehensive_plan(config)result = plan.execute()特長:
このPlanモード作成スキルは、SO8Tプロジェクトの複雑なAI開発ワークフローを効率的にPlan化し、信頼性の高い自動実行環境を構築します。
Implement comprehensive mathematical theorem proving capabilities with SFT+GRPO training, MCP/A2A agent integration, and imatrix quantization protection to surpass Boreas-phi3.5-instinct-jp in formal proof generation and scientific discovery. Use when building mathematical reasoning systems, formal verification tools, or AI-assisted theorem proving environments.
Execute GGUF quantization with imatrix protection, perform statistical benchmark evaluation with error bars, generate academic-style methodology documentation, and create comprehensive scorecards. Use when evaluating model quantization quality, comparing quantization methods, or generating publication-ready evaluation results with subagent execution and PowerShell progress visualization.
Advanced multimodal 'thinking' model pipeline with SO8T Grand Design, stability-constrained evolutionary training (EvoFreeze), and specialized OSINT/Military/Bio data collection.
Implement comprehensive mathematical theorem proving capabilities with SFT+GRPO training, MCP/A2A agent integration, and imatrix quantization protection to surpass Boreas-phi3.5-instinct-jp in formal proof generation and scientific discovery. Use when building mathematical reasoning systems, formal verification tools, or AI-assisted theorem proving environments.
Execute GGUF quantization with imatrix protection, perform statistical benchmark evaluation with error bars, generate academic-style methodology documentation, and create comprehensive scorecards. Use when evaluating model quantization quality, comparing quantization methods, or generating publication-ready evaluation results with subagent execution and PowerShell progress visualization.
Create, manage, and update multi-step execution plans for complex tasks in Cursor. Use when the user asks for a plan/roadmap/phases/checklist, or when work has multiple dependent steps, higher risk, or requires coordination. Skip for straightforward tasks. Integrates with Cursor's todo_write tool for plan tracking.