gitops-homelab

Strategy-layer agent planner. Analyses a complex request and recommends which subagents and skills to involve, in what order, with what hand-off. Use only for genuinely multi-domain tasks where the default delegation isn't obvious.

Design and build LLM-powered applications — RAG systems, agentic workflows, vector search, prompt pipelines, tool-using agents. Use for shipping production AI features, chatbots, or any AI-driven feature that goes beyond a one-shot prompt.

Generate developer-first API documentation: OpenAPI 3.1 specs, multi-language code examples, Postman collections, auth guides, and error references. Use when documenting a new API, adding endpoints, or when the user asks for OpenAPI/Postman/docs output.

Design scalable, secure, cost-efficient cloud infrastructure on AWS, Azure, or GCP. Terraform-first, FinOps-aware, multi-cloud and serverless patterns. Use for infrastructure planning, cost-reduction analysis, or cloud migration strategy.

Design context-management systems for AI-powered products — RAG, vector stores, knowledge graphs, episodic/semantic memory, multi-agent context handoff. Use when architecting AI features that need long-running or cross-session memory. Not for conversational context inside Cursor itself (handled by built-in compaction).

Design and build data-intensive systems — ETL/ELT pipelines, warehouses, streaming architectures. Expertise in Spark, Airflow, Kafka, dimensional modelling, governance, and cost optimisation. Use for new data solutions, optimising pipelines, or troubleshooting data infrastructure.

Design and write comprehensive software documentation (README, architecture, API, runbooks, ADRs, user guides). Use when creating new docs, restructuring existing docs, or tuning docs for a specific audience (developer, operator, end user, stakeholder).

Design Kubernetes platforms — GitOps (ArgoCD/Flux), service mesh, progressive delivery, multi-tenancy, security, and FinOps. Use for K8s architecture, GitOps implementation, platform engineering, or cloud-native platform design on EKS/AKS/GKE.

Generate clean, professional Mermaid diagrams (flowcharts, sequence, ERD, state, architecture, journey, gantt). Use when the user asks for a diagram, visual documentation, or to illustrate a process, flow, or system.

Build and run production ML systems — training pipelines, model serving, feature stores, monitoring, retraining. Use for deploying or operating ML models, setting up MLOps, or when a model needs to graduate from notebook to production.