# RushDB Agent Skills Agent Skills that teach AI agents to use RushDB efficiently — querying, data modeling, and persistent memory. Compatible with Claude, GitHub Copilot, Cursor, Windsurf, and any [Agent Skills](https://agentskills.io)-compatible client. --- ## Install ```bash npx skills add rush-db/rushdb --path packages/skills ``` Or install from npm: ```bash npm install @rushdb/skills ``` --- ## Available Skills | Skill | What it teaches | | --------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------- | | [`rushdb-query-builder`](#rushdb-query-builder) | Discovery-first workflow, SearchQuery syntax, aggregation, relationship traversal | | [`rushdb-agent-memory`](#rushdb-agent-memory) | Route memory across native providers, MCP, and custom harnesses without duplicate writes | | [`rushdb-memory-adapter`](#rushdb-memory-adapter) | Build and validate lifecycle-aware RushDB integrations for agent runtimes | | [`rushdb-data-modeling`](#rushdb-data-modeling) | LMPG model, label/property/relationship design, nested JSON import | | [`rushdb-faceted-search`](#rushdb-faceted-search) | Build faceted filter UIs — discover properties, enumerate values, map to widgets, assemble `where` | | [`rushdb-domain-template`](#rushdb-domain-template) | Design a tailored schema for any domain through guided conversation — interview → labels + relationships + bootstrap payload | --- ### `rushdb-query-builder` Teaches the mandatory 3-step workflow for querying RushDB: schema discovery → intent classification → query construction. Covers the full SearchQuery shape — `where` filters, `aggregate` functions, `groupBy` modes, relationship traversal, datetime operators, and vector similarity. Includes a bundled reference file (`references/search-query-spec.md`) with the complete operator reference, syntax rules, and annotated examples loaded on demand. **Triggers when an agent needs to:** - List, filter, or search records - Count, sum, average, or group data - Traverse relationships between record types - Run semantic/vector similarity search - Build any `findRecords` query --- ### `rushdb-agent-memory` Teaches how to select and operate the correct RushDB memory layer: native OpenClaw/Hermes lifecycle integration, native + MCP, MCP only, or a custom harness. It routes each read and write to one owner, enforces canonical scope boundaries, and prevents duplicate turn persistence. Includes references for integration-mode selection, AgentMemoryEvent v1, host capabilities, and operational/domain memory patterns. **Triggers when an agent needs to:** - Store session data or conversation context - Recall past decisions or prior context - Build an entity graph that survives across sessions - Search memory by meaning (semantic recall) - Associate memories via relationships --- ### `rushdb-memory-adapter` Guides contributors through building native RushDB memory support for another agent runtime: inspect exact lifecycle hooks, derive trusted authorization scope, reuse AgentMemoryEvent v1, implement fail-open recall and durable outbox delivery, and validate lifecycle conformance. **Triggers when an agent needs to:** - Implement or review an agent memory plugin/provider - Map prompt, turn, compaction, session, and shutdown hooks - Enforce participant, profile, sandbox, and subagent isolation - Add deterministic idempotency, retries, replay, and recent-write fallback - Validate a runtime adapter against the shared contract --- ### `rushdb-data-modeling` Teaches RushDB's property-centric LMPG (Labels, Multi-Properties, Graph) model: label naming conventions, property types, relationship design, how nested JSON auto-creates relationships on import, and schema evolution patterns. **Triggers when an agent needs to:** - Design a schema or data model - Choose labels and property names - Understand how nested JSON is imported - Plan relationship structure - Evolve or migrate an existing schema --- ### `rushdb-faceted-search` Teaches the full workflow for generating faceted filter UIs: call `getSchema` for structured property metadata (including `id` values), call `propertyValues` to enumerate distinct options per property, map each property type to the right UI widget (checkbox list, range slider, date picker, toggle), and assemble a live `where` clause as filters change. Covers the eager/lazy loading split, context-aware vs. original value enumeration, the two-hook pattern used in real RushDB apps, active filter chips, and full reset. **Triggers when an agent needs to:** - Generate a filter sidebar, faceted search UI, or drill-down panel - Enumerate available values for a property - Map property types to UI controls - Build a `where` clause that updates as the user selects filters - Generate `useProperties` / `usePropertyValues` style hooks --- ### `rushdb-domain-template` Guides users through designing a tailored RushDB schema via a structured interview (5 questions), then outputs a schema summary, a ready-to-run `bulkCreateRecords` bootstrap payload, and starter queries. Includes compact template cards for 10 common domains (e-commerce, SaaS/DevOps, CRM, healthcare, fintech, content, agent memory, project management, scientific research, and product management) — all adapted to RushDB's record-centric LMPG model. **Triggers when a user:** - Is starting a new project and doesn't know what labels to define - Asks "what records should I create for X?" - Wants to get up and running quickly with a known domain - Needs a starting point they can customise rather than designing from scratch --- ## Skill Structure Each skill follows the [Agent Skills](https://agentskills.io) format: ``` skills// ├── SKILL.md # Required: YAML frontmatter + instructions ├── agents/ # Optional product-facing metadata ├── references/ # Optional detailed guidance loaded on demand └── scripts/ # Optional deterministic helpers ``` --- ## License Apache 2.0