DSPy compositional prompt optimization with categorical signatures, module chaining, and automated prompt tuning. Use when building declarative LLM programs with typed signatures, composing multi-step reasoning modules (ChainOfThought, ReAct, ProgramOfThought), optimizing prompts with MIPROv2/BootstrapFewShot, or creating modular AI pipelines that separate program logic from prompt engineering.
Training manager for Hugging Face Jobs - launch fine-tuning on HF cloud GPUs with optional WandB monitoring
Domain-specific language for categorical prompt composition with functor combinators and natural transformation operators. Use when building composable prompt templates, implementing typed prompt algebras, creating reusable prompt patterns with categorical structure, or designing prompt DSLs that preserve composition properties.
Create a new Claude Code agent skill with proper folder structure and SKILL.md format. Use when the user wants to add a model-invoked skill that Claude uses autonomously based on context. Handles skill folder creation with SKILL.md and optional reference files.
Standardized reference documentation section structure for agents - project guidance, conventions, related agents, and Skills. Use when implementing or updating agent documentation.
Full categorical framework for meta-prompting with verified laws. Use when implementing functor-based task routing, monadic iterative refinement, comonadic context extraction, or quality-tracked composition. Integrates F (Functor), M (Monad), W (Comonad), and [0,1]-enriched quality categories.
Qwen Training Data Miner (Prototype)
AIシステムへのプロンプトインジェクション攻撃を防ぎ、入力検証とコンテキスト分離の設計指針を提供するスキル。 Anchors: • OWASP LLM Top 10 / 適用: LLMセキュリティ脅威モデリング / 目的: インジェクション攻撃の分類と防御パターン理解 • Simon Willison's Prompt Injection Research / 適用: 実攻撃パターン分析 / 目的: 実世界の攻撃事例から防御戦略を導出 • Defense in Depth principle / 適用: 多層防御設計 / 目的: 単一障害点の排除 Trigger: Use when designing prompt injection defenses, implementing AI security measures, sanitizing user inputs for LLM systems, separating trusted and untrusted contexts, conducting security reviews for LLM applications, mitigating indirect prompt injection risks.
AI Elements component library for AI-native applications. Use when building chatbots, AI workflows, or integrating with Vercel AI SDK's useChat hook.
AIのハルシネーション（幻覚・誤情報生成）を防止するスキル。プロンプトレベル、パラメータレベル、検証レベルの3層防御により、信頼性の高いAI出力を実現。 Anchors: • Thinking, Fast and Slow (Daniel Kahneman) / 適用: System 1/2思考分離 / 目的: 直感的推測の抑制と論理的検証の強制 • Design by Contract (Bertrand Meyer) / 適用: 入出力契約設計 / 目的: 事前条件・事後条件・不変条件による出力保証 • The Pragmatic Programmer / 適用: 実践的改善と品質維持 / 目的: 段階的検証とフィードバックループ構築 Trigger: Use when preventing AI hallucinations, ensuring factual accuracy, requiring verifiable outputs, or implementing truth constraints. hallucination prevention, fact-checking, verification, accuracy, factual output, citation required, ground truth, temperature tuning
Spivak-Niu polynomial functor implementations for learner composition and dynamical systems. Use when modeling learning systems as categorical structures, composing machine learning components with polynomial functors, implementing the categorical framework from Polynomial Functors - A Mathematical Theory of Interaction, or building compositional dynamical systems with lenses and charts.
Training manager for local GPU training - validate CUDA, manage GPU selection, monitor progress, handle checkpoints
ARIMA, SARIMA, Prophet, trend analysis, seasonality detection, anomaly detection, and forecasting methods. Use for time-based predictions, demand forecasting, or temporal pattern analysis.
Validate AI/ML models and datasets for bias, fairness, and ethical concerns. Use when auditing AI systems for ethical compliance, fairness assessment, or bias detection. Trigger with phrases like "evaluate model fairness", "check for bias", or "validate AI ethics".
Learn ADK concepts and understand the retail agent pipeline. Use when asking how the agent works, what ADK is, wanting to understand the architecture, state flow, or individual agent purposes.
This skill should be used when the user asks to "design a presentation", "structure presentation content", "improve presentation flow", "create presentation outline", "make slides more engaging", or needs guidance on storytelling, visual hierarchy, audience engagement, or presentation best practices.
Three-phase task execution framework with refinement strategies and role-specific prompting. Automatically activates when designing prompts, optimizing task flows, or implementing iterative improvement patterns.
Creates and edits Mermaid diagrams for flowcharts, sequence diagrams, ERDs, state machines, architecture diagrams, process flows, timelines, and more. Use when user mentions diagram, flowchart, mermaid, visualize, refactor diagram, sequence diagram, ERD, architecture diagram, process flow, state machine, or needs visual documentation.
Guide for authoring output-styles that transform Claude's behavior and personality. Use when creating, writing, designing, building, reviewing, or improving output-styles, persona modes, or behavior modifications. Helps with style files, keep-coding-instructions decisions, persona design, and choosing between output-styles, agents, and skills. Expert in style best practices.
Build professional image generation prompts for Nano Banana (Gemini 2.5 Flash Image). Guides users through an interactive Q&A to craft prompts using photography and cinematography terminology. Optimized for placing real photos of people into creative scenes. No API needed - outputs copy-paste ready prompts.
Capture a writer's voice DNA through collaborative interview and sample analysis. Use when someone wants to document their writing voice for use with a ghost writer skill. Produces a Voice DNA Document with patterns, anti-patterns, and actionable guidance. Handles one register/mode per session, supports refinement over time.
