Agent Skills
RTIE agent skills are procedural instruction sets that encode platform-specific knowledge for AI agents. Each skill teaches an AI assistant how to work with a specific area of the platform — the right APIs, the right constraints, and the common pitfalls to avoid.
Skills are installed automatically with the Agent Plugin, or you can download individual SKILL.md files and drop them into any AI coding tool that supports Claude's skill format.
9 skills
API key setup, MCP connection, first room, SDK install, and first chat message.
@rtie/module-sdk, manifest authoring, typed McpToolDefinition entries, context contracts, module registration.
Drizzle ORM patterns, bun db:migrate, context_id requirement, timestamptz, FK indexes, cursor pagination.
SpacetimeDB via @rtie/sdk — presence, chat, game state, manifest.spacetime tables. Never direct SDK calls.
Provider-neutral RTIE actor model, API keys, BYOA/OIDC adapters, hosted auth, session tokens, workspace_type gating, tenant isolation, role model.
Signal ingestion, rule authoring (condition operators, action types), <350ms end-to-end dispatch, action log.
Browse listings, monitor live auctions, place bids and buy-now with human confirmation. Buyer agent patterns.
Endpoint registration, HMAC signature verification, event catalog, retry policy, idempotency patterns.
wrangler deploy, CF Pages, environment secrets, HyperDrive setup, Turborepo affected builds, local dev.
Adding skills manually
Download a SKILL.md file and place it anywhere in your project. Claude Code and other Claude-powered agents will pick up skills from .claude/skills/ or from paths listed in your plugin config.
Skills follow the Claude agent skill format: a YAML frontmatter block (name, description, version) followed by markdown instructions.