Runtype Build Product

Runtype Build Product

Use this skill when the user wants to create or materially change a Runtype product. MCP is the source of truth for schemas, catalogs, and current platform rules.

Required First Calls

Before designing or creating resources:

• Product build: get_build_instructions(task="build-product", description=...).
• Flow build: get_build_instructions(task="generate-flow", description=..., name=...).
• Capability scoping: get_build_instructions(task="explain-capabilities").

Then fetch only relevant get_platform_documentation topics:

Build instructions route deeper subjects such as skills, subagents, retrieval, commerce, and evals. Do not fetch every catalog in advance or read the same topic again as a resource.

Design Policy

Answer these before building:

• Who is the product for: personal, internal team, external customers, machines, or agents?
• Where do users work: website chat, Slack, email, SMS, messaging apps, webhook, API, MCP, schedule, or A2A?
• What is the capability: agent, flow, existing agent, existing flow, or external agent?
• What tools and state does it need: built-ins, Orthogonal tools, MCP servers, external HTTP tools, custom code, local SDK tools, WebMCP page tools, records, secrets?

Start with an agent unless the work is a fixed sequence. Use flows for deterministic pipelines, indexing, batch processing, artifact rendering, and hot paths that should be cheap and fast. Use subagents or multiple capabilities instead of one giant prompt.

When the product needs a custom tool (an external HTTP tool, custom code, or a flow exposed as a tool), design it with the tool-design skill family before create_tool: it covers the model-facing description and schema, result shape, error classes, idempotency, secrets, and the save-time validator codes. If that skill is not installed, the same checklist is public at https://github.com/runtypelabs/skills/tree/main/skills/tool-design.

For commerce products, check UCP support when a merchant domain is known, summarize the finding, and ask whether to use UCP or the traditional commerce path before proceeding.

Build Loop

1. Discover only the account state the task needs with get_me, list_products, list_agents, list_flows, list_tools, list_model_configs, and product-scoped list_surfaces when relevant. These large inventory tools use compact string previews by default; keep that shape for discovery, use agent_type when narrowing agents, and call the matching get_* tool for complete configuration. Request view: "full" only when complete strings are needed for every row.
2. Pick surfaces from the live surface-types docs. Include messaging as the generic multi-channel surface, and prefer dedicated surfaces when channel constraints matter. For browser-side WebMCP tools, use a chat surface with behavior.webmcp rather than an mcp surface. For the WebMCP mechanics and the advanced non-Persona raw-/v1/dispatch clientTools[] path, use runtype-persona.
3. Prefer first-party/built-in tools, then Orthogonal tools, then MCP servers, then custom HTTP/JS tools.
4. Validate before creating: validate_product, validate_flow, validate_product_flow, validate_product_agent, validate_product_surface, validate_product_tool, and validate_code for custom JS or transform code.
5. create_product creates an empty container, not an FPO import. Create the required agents/flows and tools, attach capabilities and surfaces using returned IDs, and create schedules or credentials only when requested delivery requires them. Read get_product_setup before testing integrations; complete authorized setup or return the remaining human-action links. A bare flow does not need product setup.
6. Test at the user-facing layer. Use execute_agent, dispatch, execute_tool, run_flow, submit_batch, submit_eval, trace_execution, and trace_conversation as appropriate. Use fixtures/sandbox targets for side effects; do not send live messages, charge money, or activate recurring work merely to smoke-test. Report validation, setup readiness, tested behavior, and untested paths separately.

Guardrails

• Never invent schemas or model IDs; fetch docs and model configs.
• Do not inline credentials. Use {{secret:KEY}} and secret intake.
• Read before update; preserve fields the user did not ask to change.
• Check update semantics in the live schema; preserve sibling config fields when replacing nested objects. update_flow.steps replaces the entire step list.
• Surface-level evals catch orchestration and formatting issues that per-agent evals miss.
• If a product needs deeper or newer platform rules than this skill names, fetch platform-catalog and focused direct resources instead of appending feature notes here.

If the umbrella runtype skill is installed alongside this focused skill, its durable references provide deeper fallback notes. This skill must still work when installed by itself; prefer live MCP docs over local sibling files.