Browse reusable Agent Skills, each with a clear purpose and practical guidance.
Expert LLM architect specializing in large language model architecture, deployment, and optimization. Masters LLM system design, fine-tuning strategies, and production serving with focus on building scalable, efficient, and safe LLM applications.
LLMアプリケーションの設計・運用・評価・最適化。プロンプト管理、RAG構築、ファインチューニング、評価パイプライン、コスト最適化、本番運用。「LLM」「プロンプト」「RAG」「ファインチューニング」「AI運用」「評価」に関する質問で使用。
Creates missing instruction files (CLAUDE.md, AGENTS.md, GEMINI.md) and audits all for token budget, prompt cache safety, and cross-agent consistency. Use after setup or when instruction files need alignment.
Generates individual .claude/commands files from templates
LocalSend-based P2P transfer with MCP server design for NATS/Tailscale discovery and throughput tuning.
Agentic PRD batch: when user says 'process PRDs' or 'run batch from PRD folder'; implement per spec; update status; write outcome; reference PRD schema + external-agent guide. Use when editing .agent/queue/**, prds/**, *.prd.md.
Umbrella/queue: only when queue or PRD in scope; named workflows; impact summary + user confirmation before big/destructive runs. Use when editing .agent/queue/**, .agent/**, PRD.md, *.prd.md or when user invokes a named umbrella workflow.
Agent Q bidirectional transport client - file_drop ship/ingest, mail pull/ship (including strict gpg preencrypted), registry validation, queue-pending, archive-prune. Use when shipping or ingesting sealed PRD manifests between agents over shared folders or mail; when editing agent_trust_registry, manifest.schema.json, or agentq CLI.
Decision tree / reverse prompt workflow - AI prompts user one question at a time with 4 options (A-D), preferred choice + rationale; build context for maximum impact. Use when user says 'decision tree', 'run the decision tree', 'reverse prompt', or 'reverse prompt workflow'; or when editing .agent/queue/**, PRD.md, *.prd.md.
Use when preparing to delegate to MacEff subagents. Read policies to discover current delegation patterns through timeless questions that extract details without prescribing answers.