Expert capability builder for Sunrise orchestration. Creates custom agent capabilities (tools) that let agents call APIs, look up data, process refunds, send notifications, or perform any external action. Handles Zod validation, OpenAI-compatible function definitions, execution handlers, registry wiring, and database setup via the 4-step pipeline: TypeScript class, registry, DB row, agent binding. Use when an agent needs a new tool, needs to call an external service, or needs to perform actions beyond conversation.
End-to-end orchestration solution builder for Sunrise. Takes a business problem — "build me a customer support chatbot", "I need an AI assistant that can look up orders and process refunds", "create a content review pipeline" — and produces everything needed: providers, agents, capabilities, knowledge base, and workflows. Handles the full implementation pipeline from problem description to running system. Use when building complete agentic solutions, implementing an architecture from agent-architect, or setting up a new agent system from scratch.
Testing lens and quick patterns reference for Sunrise. Apply this skill whenever you create, modify, or audit Vitest tests — unit, integration, or component. Its purpose is to enforce an anti-green-bar mindset (tests must verify what the code *does*, not what mocks return) and route work into the right tool: `/test-plan` → `/test-write` → `/test-review` → `/test-fix` for branch- or module-scoped work, the `test-engineer` agent for spawn-based writing, or hands-on patterns for one-off tests. Use when the user asks to add, write, audit, fix, or review tests.
Architect for Sunrise's agent orchestration system. Designs agentic solutions by selecting from 21 design patterns, composing multi-pattern architectures, and mapping designs to Sunrise's orchestration primitives (agents, capabilities, workflows, knowledge bases). Use when a developer wants to design, plan, or debug an AI agent system — whether they say "build me a chatbot", "I need an agent that can look up orders", "design an AI pipeline", or "why is my agent hallucinating". This skill handles the DESIGN phase; the orchestration-solution-builder skill handles IMPLEMENTATION.
Expert knowledge base builder for Sunrise orchestration. Sets up document ingestion, chunking, embeddings, and vector search so agents can answer questions grounded in real data instead of hallucinating. Handles the full lifecycle: upload documents (MD, PDF, EPUB, DOCX), chunk, generate embeddings, and scope knowledge to specific agents. Use when agents need to search company docs, product information, FAQs, or any document corpus.
Expert workflow builder for Sunrise orchestration. Composes multi-step agent pipelines as workflow DAGs — routing requests to different agents, chaining LLM calls, adding human approval gates, running parallel branches, integrating RAG retrieval, and bolting on post-hoc supervisor audit + deterministic Markdown reporting. Uses 19 step types, template interpolation, error strategies, run-time toggles, and budget enforcement. Use when building multi-step agent pipelines, adding approval flows, connecting multiple agents in a sequence, or adding an honest end-of-workflow audit.
Deployment guide generator for Sunrise. Creates platform-specific deployment documentation with environment variable checklists, database setup, and troubleshooting guides. Use when deploying to new platforms or updating deployment docs.
Documentation writer for Sunrise's .context/ substrate. Creates and updates documentation following established patterns: actionable over descriptive, project-specific quirks, anti-patterns first, verified against actual code. Use when adding documentation for new features or updating existing docs.