constructive-skills

Declarative schema definition — blueprints, node type registry, presets, and the blueprint definition format. Use when asked to 'create a blueprint', 'blueprint template', 'construct blueprint', 'copy template', 'blueprint definition', 'definition hash', 'table hashes', 'schema marketplace', 'blueprint provisioning', 'module presets', 'what modules should I install', 'auth:email preset', 'b2b preset', or when working with blueprint JSONB definitions.

Feature catalog for Constructive apps — one scannable index of every capability (auth methods, authorization, multi-tenancy, data modeling, storage, search, AI, GraphQL, ops) with pointers to the authoritative skill or toggle that owns it. Use when asked "what can a Constructive app do", "what features are available", "which skill covers X", "is there built-in support for Y", "how do I enable read-only mode / magic links / passkeys / SSO / pgvector / search / uploads", or when orienting to the platform for the first time.

Authorization — Safegres protocol, 23 Authz* policy types, RLS, grants, permissions, permission defaults, GuardStepUp, read-only access, storage policies, and the secureTableProvision workflow. Use when asked to 'add security', 'RLS', 'grants', 'policies', 'Safegres', 'Authz*', 'AuthzEntityMembership', 'AuthzDirectOwner', 'AuthzMemberOwner', 'AuthzComposite', 'AuthzSystemOnly', 'AuthzHumanOnly', 'AuthzValueAllowed', 'AuthzValueExists', 'AuthzValueMatch', 'system-only policy', 'system-only writes', 'human-only mutation', 'block agents from writing', 'read-only mode', 'secure table provision', 'storage policies', 'bucket security', 'permission model', 'permission defaults', 'default_permissions', 'GuardStepUp', 'step-up auth', 'guard step-up', 'require step-up', 'MFA guard', 'named permissions', or when working with authorization in blueprints or the ORM.

Build a Constructive app end-to-end from a brief. Use when asked to build/scaffold an app on Constructive, create a CRUD app, provision a data model with RLS, wire Blocks + auth flows, or verify an app with Chrome QA / agent-browser — keep time-to-app under 10 minutes on a warm hub. Drives the four mainline phases (1 Backend Up, 2 Data Model Provisioned, 3 Frontend + SDK, 4 UI / Blocks) and supports the three policy tiers: owner-scoped, public-read + owner-write, and b2b org-membership. Triggers: "build a Constructive app", "scaffold an app on Constructive", "provision a data model with RLS", "wire Blocks and auth flows", "verify the app with agent-browser". Builds ANY CRUD app generically from a brief — it is NOT for editing the Constructive platform itself (constructive / constructive-db).

AI — agent module, LLM providers, RAG pipelines, embeddings, agentic-kit multi-provider client, Search* blueprint nodes (SearchUnified, SearchVector), ProcessFileEmbedding/ProcessImageEmbedding/ProcessChunks. Use when asked to 'add AI search', 'build RAG pipeline', 'embedding worker', 'agentic-kit', 'LLM integration', 'Ollama', 'Anthropic', 'OpenAI', 'file embedding', 'image embedding', 'chunking', 'SearchUnified', 'SearchVector', 'multiplayer agents', 'shared agents', 'multi-agent', 'agent_id', or when working with AI features in blueprints.

Identity, login, sessions, MFA, devices, identity linking — auth flow (sign-up/sign-in/JWT), auth settings (MFA, anonymous sessions, CAPTCHA, cookie auth, rate limits), device settings (tracking, trusted devices, approval gate), identity linking (link_identity, account collision resolution, allow_link_by_email, primary auth method), and service-level auth config. Use when asked to 'sign up', 'sign in', 'auth flow', 'MFA', 'magic link', 'passkeys', 'device approval', 'trusted devices', 'session management', 'cookie auth', 'anonymous sessions', 'CAPTCHA', 'rate limits', 'bootstrap user', 'JWT', 'link identity', 'identity linking', 'account collision', 'allow_link_by_email', 'primary auth method', 'connected accounts', or when working with authentication.

Principals — scoped sub-identities for API keys and agents. A principal is a delegated identity owned by a human (or org) that authenticates via an API key and operates with a subset of its parent's permissions. Use when asked to 'create an API key', 'issue an agent credential', 'scope an agent to an org', 'read-only API key', 'revoke an API key', 'create a principal', 'org API key', 'service account', 'machine identity', 'agent identity', 'bypass step-up for a bot', 'principalEntity', 'principalScopeOverride', or when managing agent/API-key identities via the SDK ORM.

Subscriptions, notifications, change_log, CursorTracker, ephemeral realtime — real-time data via GraphQL subscriptions, PostGraphile live queries, change tracking, and ephemeral pg_notify delivery. Use when asked to 'add subscriptions', 'real-time updates', 'live queries', 'notifications', 'change_log', 'CursorTracker', 'websocket', 'ephemeral realtime', 'pg_notify', 'high-frequency signals', or when working with real-time features in Constructive.

Build Constructive frontend UIs — 50+ components (@constructive-io/ui on Base UI + Tailwind CSS v4), CRUD Stack cards (iOS-style slide-in panels), and dynamic _meta forms (zero-config CRUD for any table). Use when building UIs, creating forms, working with Stack cards, or using the component library.

Generated ORM — query patterns, mutations, relations, pagination, _meta introspection, runtime query building, and query keys. Use when asked to 'query data', 'ORM patterns', 'findMany', 'findOne', 'create', 'update', 'delete', 'paginate', 'cursor pagination', 'offset pagination', '_meta', 'runtime query', 'query builder', or when using the generated @constructive-io/sdk ORM client.

All search strategies — tsvector full-text, BM25, trigram, pgvector semantic, PostGIS spatial, unified composite search, and RAG patterns. Use when asked to 'add search', 'full-text search', 'tsvector', 'BM25', 'trigram', 'pgvector', 'PostGIS', 'unified search', 'searchScore', 'unifiedSearch', 'semantic search', 'vector search', 'spatial search', 'RAG search', or when working with Search* blueprint nodes or graphile-search adapters.

Notification system — inbox, delivery, channels, preferences, bounce/complaint handling, suppression list, and digest batching. Use when asked to 'add notifications', 'notification preferences', 'bounce handling', 'complaint handling', 'suppression list', 'delivery log', 'notification channels', 'digest batching', 'quiet hours', 'webhook ingestion', 'SES bounces', 'SendGrid webhooks', 'channel deactivation', or when working with notifications_module in blueprints.

Access control — roles, permissions, profiles, grants, membership access, and entity-scoped authorization. Use when asked to 'assign permissions', 'create roles', 'set up profiles', 'grant access', 'permission defaults', 'admin vs owner', 'membership permissions', 'effective permissions', 'revoke access', 'role hierarchy', 'custom roles', 'profile bundles', 'default profile', 'entity-scoped permissions', 'org permissions', 'app permissions', 'permission resolution', or when working with the access control model in blueprints or the ORM.

Multi-tenancy, memberships, invites, entity types, entity-scoped storage, agent modules, and namespace modules. Use when asked to 'create entity types', 'add channels/teams/data rooms', 'provision entity storage', 'entity-scoped buckets', 'agent_module', 'namespace_module', 'invite users', 'profile assignment', 'memberships', 'multi-tenancy', 'org-scope', 'app-scope', or when working with entity_types in blueprints.

Billing, limits, plans, credits, feature flags, meters, and usage tracking. Use when asked to 'set up billing', 'create meters', 'grant credits', 'record usage', 'check quota', 'universal credits', 'billing provider', 'set up limits', 'feature flags', 'cap tables', 'LimitCounter', 'LimitAggregate', 'LimitFeatureFlag', 'apply plan', 'transfer quota', 'credit expiration', 'rollover credits', or when working with billing or limits in blueprints.

Background job system — JobTrigger blueprint node for enqueuing jobs on row changes (with compound conditions support: AND/OR/NOT combinators, column-aware type resolution), ProcessFileEmbedding/ProcessImageEmbedding/ProcessChunks composition wrappers, payload strategies, the Knative worker pipeline, scheduled jobs, and the app_jobs database extension. Use when asked to 'trigger a job', 'enqueue a background task', 'add a job trigger', 'run a function on row change', 'schedule a job', 'compound conditions', 'file embedding trigger', 'image embedding trigger', 'multi-modal embedding', or when working with JobTrigger/ProcessFileEmbedding/ProcessImageEmbedding/ProcessChunks in blueprints.

Code generation pipeline — config, templates, AST transforms, schema introspection, and generating typed TypeScript clients (ORM, hooks, CLI) from GraphQL schemas. Use when asked to 'generate code', 'codegen', 'generate ORM', 'generate hooks', 'generate CLI', 'cnc codegen', 'graphql-codegen config', 'schema export', 'introspection', or when working with @constructive-io/graphql-codegen.

Tables, fields, relations, constraints, indexes, enums, and database provisioning via the type-safe SDK. Use when asked to 'create a table', 'add a field', 'add a column', 'create a relation', 'add a constraint', 'add an index', 'create a foreign key', 'define field types', 'provision a database', 'create an enum', 'api_required', or when working with metaschema_public operations.

EventTracker, achievements, referrals, invite virality — EventTracker blueprint node for recording events on row changes, achievements[] for levels with credit rewards, EventReferral for multi-level referral chains, period-aware counting. Use when asked to 'add analytics', 'track events', 'add achievements', 'gamification', 'EventTracker', 'level requirements', 'achievement rewards', 'invite virality', 'referral credits', 'EventReferral', 'max_depth', 'multi-level referral', or when working with events_module in blueprints.

Graph module and merkle store — SDK-authorable graph infrastructure (graphs, executions, outputs) with entity scoping, content-addressed merkle state tracking (objects, stores, commits, refs), and FBP integration. Use when asked to 'create a graph', 'graph module', 'merkle store', 'content-addressed storage', 'flow-based programming', 'graph executions', 'FBP', or when working with graph_module or merkle_store_module in blueprints.

Generated React Query hooks — query/mutation hooks, cache management, optimistic updates, and hook patterns. Use when asked to 'use hooks', 'React Query hooks', 'useQuery', 'useMutation', 'cache invalidation', 'optimistic updates', 'generated hooks', or when working with the generated React Query client from @constructive-io/graphql-codegen.

Internationalization — DataI18n translation tables, multilingual tsvector search with per-row language stemming (30+ languages), i18n_module for app-level language config, lang_column dynamic stemming, and graphile-i18n plugin for Accept-Language negotiation. Use when asked to 'add translations', 'make fields translatable', 'multilingual search', 'i18n', 'DataI18n', 'i18n_module', 'lang_column', 'per-language stemming', 'translation tables', 'locale strings', 'Accept-Language', or when working with i18n in blueprints.

Platform core — services/schemas, deployment, server config, cloud functions, environment configuration, and the cnc CLI execution engine. Use when asked to 'deploy constructive', 'start the server', 'cnc server', 'GraphQL API', 'configure domains', 'cloud functions', 'Knative', 'env config', 'cnc execute', 'docker compose', 'set up service', 'API routing', 'services_public', or when working with server deployment or environment configuration.

Uploads, buckets, presigned URLs, file lifecycle — presigned URL flow (requestUploadUrl → PUT → downloadUrl), bucket provisioning, upload-client, public/private/entity-scoped buckets, MIME type restrictions, file size limits, and deduplication. Use when asked to 'upload files', 'add file uploads', 'configure storage', 'set up MinIO', 'presigned URLs', 'download URLs', 'bucket provisioning', or when working with storage plugins.