@joktec/skills 0.1.3 → 0.1.7

package/dist/codex/skills/joktec-database-extended-skill/SKILL.md

@@ -23,6 +23,7 @@ Use this skill for database clients that are not Mongo or MySQL.
 - Use package services for client lifecycle and connection config.
 - Do not claim parity with Mongo/MySQL repository behavior unless the package implements it.
 - Use package README and source as final truth before adding advanced behavior.
+- If guidance is insufficient, read this skill's references and inspect `../joktec-framework` package source or GitHub fallback before assuming APIs.
 
 ## Reference
 

package/dist/codex/skills/joktec-database-extended-skill/references/extended-databases.md

@@ -1,5 +1,15 @@
 # Extended Database Usage
 
+## Source Lookup
+
+When blocked, inspect:
+
+- `packages/databases/README.md`
+- `packages/databases/AGENTS.md`
+- `packages/databases/<package>/README.md`
+- `packages/databases/<package>/src/index.ts`
+- package config/module/service files under `src/`
+
 ## Package Boundaries
 
 Extended database packages expose reusable clients and helpers for specific data systems. They should not contain consumer app schemas or business-specific query policy.
@@ -9,3 +19,17 @@ Extended database packages expose reusable clients and helpers for specific data
 - Elastic may depend on `@joktec/http` for HTTP-backed behavior.
 - Arango and BigQuery are separate client packages; verify the README/source before assuming repository-like CRUD support.
 - Use local package tests or consumer harnesses only when the target service stack is available.
+
+## Best Practices
+
+- Treat these packages as client wrappers unless source explicitly exposes a repository abstraction.
+- Keep index names, dataset names, query templates, and domain-specific projections in the consumer app.
+- Validate credentials and endpoints through config; never commit service account data.
+- Keep retries, timeouts, and debug logging visible in config.
+- Use provider-specific tests/mocks for package logic and live stack tests only when credentials/services are intentionally available.
+
+## Anti-Patterns
+
+- Do not apply MongoRepo/MysqlRepo assumptions to Elastic/Arango/BigQuery.
+- Do not introduce app schemas or business query policy into reusable database clients.
+- Do not claim support for a provider feature unless the package source or README shows it.

package/dist/codex/skills/joktec-framework-skill/SKILL.md

@@ -25,6 +25,7 @@ Start here when a task mentions JokTec generally, multiple `@joktec/*` packages,
 - Prefer config-driven module setup and `conId` when a package supports multiple clients.
 - Preserve NestJS module boundaries, dependency injection, lifecycle hooks, and exported package APIs.
 - Do not invent behavior for unfinished or missing packages.
+- If a focused skill is loaded without this entrypoint, still apply source-first lookup before assuming APIs.
 
 ## Reference
 

package/dist/codex/skills/joktec-framework-skill/references/framework-map.md

@@ -1,5 +1,24 @@
 # JokTec Framework Map
 
+## Source-First Operating Protocol
+
+Treat this skill pack as curated guidance, not the final implementation truth. When a task is unclear, or when a package API seems different from the skill text:
+
+1. Inspect the consumer project first to understand the package versions and local usage pattern.
+2. Prefer local framework source at `../joktec-framework` when available.
+3. Read package `README.md`, nearest `AGENTS.md`, and `src/index.ts` before changing code.
+4. If local source is unavailable, use `https://github.com/joktec/joktec-framework` as fallback.
+5. Do not invent behavior that is not visible in framework source or package docs.
+6. If package docs and source conflict, source code wins and the mismatch should be treated as documentation drift.
+
+High-value source files:
+
+- `AGENTS.md` and `docs/agents/*` for framework-level architecture and runtime policy.
+- `packages/<family>/AGENTS.md` for family boundaries.
+- `packages/<family>/<package>/README.md` for developer-facing usage.
+- `packages/<family>/<package>/src/index.ts` for public API.
+- Package config/module/service/repository files for real runtime behavior.
+
 ## Package Families
 
 - `@joktec/core`: bootstrap, config, logger, metrics, guards, base abstractions, transports, pagination, Bull.
@@ -17,9 +36,24 @@
 
 Consumer apps depend on concrete `@joktec/*` packages. Concrete packages usually depend on `@joktec/core` and `@joktec/utils`. Keep app-specific schemas, handlers, and business semantics outside reusable packages.
 
+Do not reverse the dependency direction:
+
+- reusable packages must not import from `apps/`
+- `common` packages must not depend on concrete adapters/brokers/databases
+- consumer apps own business workflows, DTO composition, schemas/entities, queue names, topics, and provider credentials
+
 ## Common Consumer App Pattern
 
 1. Register package modules in the app module or repository module.
 2. Keep app schemas/entities and app repositories in the consumer app.
 3. Extend the package repository base class when the package provides one.
 4. Use `BaseService`, `BaseController`, or transport helpers from `@joktec/core` when the app follows standard CRUD or message patterns.
+
+## Skill Selection
+
+- Use `joktec-common-skill` for BaseController/BaseService/config/pagination/client lifecycle/utils/cron.
+- Use `joktec-mongo-skill` when the resource uses Mongo schemas, MongoRepo, Typegoose decorators, plugins, or ObjectId-safe queries.
+- Use `joktec-mysql-skill` when the resource uses TypeORM entities, MysqlRepo, SQL dialect behavior, transactions, or schema-first column decorators.
+- Use broker/adapter/integration/tool skills for provider wiring; keep business semantics in the consumer app.
+
+When using the ecosystem `npx skills` installer, install `joktec-framework-skill` and `joktec-common-skill` together with any focused package skill. The current ecosystem installer does not auto-resolve dependencies from frontmatter.

package/dist/codex/skills/joktec-integration-skill/SKILL.md

@@ -22,6 +22,7 @@ Use this skill for third-party integration packages.
 - Use package services instead of direct SDK initialization in app code.
 - Preserve app-neutral service behavior; consumer apps own domain workflows.
 - Verify current package README/source before relying on advanced provider features.
+- If guidance is insufficient, read this skill's references and inspect `../joktec-framework` package source or GitHub fallback before assuming APIs.
 
 ## Reference
 

package/dist/codex/skills/joktec-integration-skill/references/integrations.md

@@ -1,9 +1,38 @@
 # Integration Usage
 
+## Source Lookup
+
+When blocked, inspect:
+
+- `packages/integrations/README.md`
+- `packages/integrations/AGENTS.md`
+- `packages/integrations/<package>/README.md`
+- `packages/integrations/<package>/src/index.ts`
+- package config/module/service files under `src/`
+
 ## Firebase
 
 Use the integration module and service to initialize Firebase Admin clients from validated config. Keep credential files local or environment-provided.
 
+Best practice:
+
+- Keep service account JSON, private keys, and project credentials out of git.
+- Prefer environment-specific config or ignored local credential files.
+- Keep notification/user/storage/product semantics in the consumer app.
+- Mock Firebase Admin SDK in package tests; use live credentials only in explicit integration environments.
+
 ## GPT
 
 Use the integration package for reusable GPT client setup. Keep prompt/business policy in the consumer app, not in the generic package.
+
+Best practice:
+
+- Keep prompts, model choice policy, tool schemas, and product safety rules in the consumer app.
+- Keep API keys in secret management or environment config.
+- Verify current package completeness before relying on advanced APIs; `@joktec/gpt` may lag behind provider SDK changes.
+
+## Anti-Patterns
+
+- Do not commit real credential files.
+- Do not encode product-specific prompts or notification logic into the reusable integration package.
+- Do not assume provider SDK behavior without checking package source and provider docs when APIs are unstable.

package/dist/codex/skills/joktec-mongo-skill/SKILL.md

@@ -17,8 +17,16 @@ Use this skill for MongoDB-backed resources that rely on JokTec's Mongoose/Typeg
 - Keep schema classes, app repositories, and app-specific queries in the consumer app.
 - Extend `MongoRepo<T, ID>` for app repositories.
 - Preserve `conId` when the app has multiple Mongo connections.
-- Use schema-first decorators when a schema class should be reused as a DTO source.
+- Use schema-first decorators when a schema class should be reused as a DTO source; wrappers should reduce repeated Typegoose, validator, transformer, and Swagger stacks.
+- Use `RefId<T>` for stored reference id fields and `PopulatedRef<T>` for populated virtual fields.
+- Use `@Schema({ kind: 'embedded' })` for value objects without `_id` or timestamps.
+- Use `@Schema({ kind: 'subdocument' })` for embedded documents that need their own `_id` and timestamps.
+- Use `@Prop({ kind: 'virtual', mode: 'getter' })` for computed getters that need expose/Swagger metadata without persistence.
+- Use `@Prop({ ref: () => Target, foreignField, localField })` for populate-one virtuals when inferred defaults are enough.
+- Use `@Prop({ type: () => [Target], ref: () => Target, foreignField, localField })` for populate-array virtuals.
+- Use `@Prop({ kind: 'map', type: Object })` for map/snapshot payloads that must keep their raw shape.
 - Treat ObjectId casting and regex behavior as safety-sensitive.
+- For real migrations, inspect `node_modules/@joktec/mongo` first, then installed README or GitHub package docs, then GitHub source before assuming APIs.
 
 ## References
 

package/dist/codex/skills/joktec-mongo-skill/references/repository.md

@@ -1,13 +1,45 @@
 # Mongo Repository Usage
 
+## Source Lookup
+
+When blocked, inspect:
+
+- Consumer project first: `node_modules/@joktec/mongo`.
+- Installed docs next: `node_modules/@joktec/mongo/README.md`.
+- GitHub docs next: `https://github.com/joktec/joktec-framework/tree/main/packages/databases/mongo`.
+- GitHub source fallback:
+  - `packages/databases/mongo/src/index.ts`
+  - `packages/databases/mongo/src/mongo.module.ts`
+  - `packages/databases/mongo/src/mongo.service.ts`
+  - `packages/databases/mongo/src/mongo.repo.ts`
+  - `packages/databases/mongo/src/helpers/mongo.helper.ts`
+  - `packages/databases/mongo/src/helpers/mongo.pipeline.ts`
+  - `packages/databases/mongo/src/helpers/mongo.utils.ts`
+  - `packages/databases/mongo/src/models/*`
+
 ## Module Setup
 
 Register schemas with `MongoModule.forRoot({ conId, models: [...] })`. Use the same `conId` in app repositories when using non-default connections.
 
+Best practice:
+
+- Register app schema classes in the consumer app repository module.
+- Use one owner process for index creation in multi-process deployments.
+- Preserve `conId` through service/repo constructors for multi-connection apps.
+- Do not rely on the global mongoose model registry when `MongoService` provides connection-aware resolution.
+
 ## Repository Pattern
 
 Extend `MongoRepo` and pass the schema class to the base constructor. Services can then use `BaseService` from `@joktec/core`.
 
+Repository checklist:
+
+- Keep schema-specific query helpers in the app repository, not in controllers.
+- Use repository methods for standard reads so query parsing, soft delete, populate, and pagination stay consistent.
+- Pass transaction/session options through read-modify-write flows when the app uses transactions.
+- Repository read paths should return schema class instances with normalized ObjectId/string values, including populated and deep-populated values.
+- Code that needs raw Mongoose documents should use `MongoService.getModel(...)` or Typegoose/Mongoose APIs directly.
+
 ## Query Safety
 
 - Root `id` can act as an API alias for `_id` in query conditions.
@@ -15,6 +47,20 @@ Extend `MongoRepo` and pass the schema class to the base constructor. Services c
 - `$like`, `$begin`, and `$end` escape regex input by default.
 - Do not rely on broad conversion when storing raw snapshots, maps, or nested subdocuments.
 
+Anti-patterns:
+
+- Do not convert every nested `id` key into `_id`; only API-facing query aliases should be converted.
+- Do not cast every 24-character hex string into ObjectId; fields like `externalId`, `code`, and `slug` may be strings.
+- Do not inject soft-delete conditions into unknown nested aggregate branches unless the target collection is known.
+
 ## Pagination
 
 `MongoRepo.paginate` supports page, offset, and cursor responses. Cursor mode defaults to `_id`; custom `cursorKey` appends `_id` as a tie-breaker.
+
+Cursor checklist:
+
+- Use `_id` for default Mongo cursor ordering.
+- Use `createdAt` or another indexed stable key when the product requires chronological cursor behavior.
+- Append `_id` as a tie-breaker for non-unique cursor keys.
+- Fetch `limit + 1` records and generate `nextCursor` only when another page exists.
+- Treat cursor tokens as opaque; clients should not parse them.

package/dist/codex/skills/joktec-mongo-skill/references/schema-and-plugins.md

@@ -1,17 +1,82 @@
 # Mongo Schema And Plugins
 
+## Source Lookup
+
+When blocked, inspect:
+
+- Consumer project first: `node_modules/@joktec/mongo`.
+- Package docs next: `node_modules/@joktec/mongo/README.md`.
+- GitHub docs next: `https://github.com/joktec/joktec-framework/tree/main/packages/databases/mongo`.
+- GitHub source fallback:
+  - `packages/databases/mongo/src/decorators/schema.decorator.ts`
+  - `packages/databases/mongo/src/decorators/schema.options.ts`
+  - `packages/databases/mongo/src/decorators/prop.decorator.ts`
+  - `packages/databases/mongo/src/decorators/props/*`
+  - `packages/databases/mongo/src/models/mongo.ref.ts`
+  - `packages/databases/mongo/src/models/object-id.ts`
+  - `packages/databases/mongo/src/models/mongo.schema.ts`
+  - `packages/databases/mongo/src/plugins/paranoid.plugin.ts`
+  - `packages/databases/mongo/src/plugins/strict-reference.plugin.ts`
+  - `packages/databases/mongo/src/plugins/transform.plugin.ts`
+
 ## Schema Decorators
 
 Use `@Schema` and `@Prop` wrappers from `@joktec/mongo` for Typegoose schema metadata plus validation, transform, and Swagger metadata.
 
 Avoid mutating shared option objects across multiple properties.
 
+Best practice:
+
+- Use schema-first decorators when the schema should be reused by DTO mapped types.
+- Keep raw `@prop` or raw Mongoose decorators only when the wrapper cannot express the behavior.
+- Pass custom validators/transforms explicitly rather than adding hidden global behavior.
+- Keep maps, snapshots, and dynamic objects explicit so helper conversion does not alter their shape.
+- Keep app-level reference semantics visible; strict reference plugin checks existence, but the app still owns domain rules.
+- Use `RefId<T>` for persisted id fields and `PopulatedRef<T>` for populated virtual instance fields.
+- Use lazy `type` resolvers such as `type: () => User` or `type: () => [User]` when the wrapper cannot infer the runtime class.
+- Use `@Schema({ kind: 'embedded' })` for value objects without `_id` or timestamps.
+- Use `@Schema({ kind: 'subdocument' })` for embedded documents that need `_id` and timestamps but should not create a collection.
+- Use `@Prop({ kind: 'virtual', mode: 'getter', comment, optional, hidden, expose, swagger })` for computed getters that only need class-transformer and Swagger metadata.
+- Use `@Prop({ ref: () => User, foreignField, localField })` for populate-one virtuals when inferred defaults are enough.
+- Use `@Prop({ type: () => [User], ref: () => User, foreignField, localField })` for populate-array virtuals.
+- Use `@Prop({ kind: 'map', type: Object })` for raw maps/snapshots instead of passing `PropType.MAP` at the call site.
+
+Common mappings:
+
+| Use case | Preferred shape |
+| --- | --- |
+| Stored single reference id | `fieldId?: RefId<User>` with `@Prop({ type: ObjectId, ref: () => User })` |
+| Stored reference id array | `fieldIds?: RefId<User>[]` with `@Prop({ type: [ObjectId], ref: () => User })` |
+| Embedded value object | `@Schema({ kind: 'embedded' })` on the nested class |
+| Embedded document | `@Schema({ kind: 'subdocument' })` on the nested class |
+| Raw map/snapshot | `@Prop({ kind: 'map', type: Object })` |
+| Populated single virtual | `field?: PopulatedRef<User>` with `@Prop({ ref: () => User, foreignField: '_id', localField: 'fieldId' })` |
+| Populated virtual array | `fields?: PopulatedRef<User>[]` with `@Prop({ type: () => [User], ref: () => User, foreignField: '_id', localField: 'fieldIds' })` |
+| Computed getter | `@Prop({ kind: 'virtual', mode: 'getter', comment: '...' }) get value() { ... }` |
+
+Populate inference:
+
+- `ref` + `localField` + `foreignField` marks the field as virtual populate.
+- Populate-one can fallback to the same class from `ref`.
+- Populate arrays still need `type: () => [Target]` because runtime reflection only sees `Array`.
+- `justOne` defaults to `true` for non-array populate fields.
+- Swagger examples default to `{}` or `[]` for populated fields unless explicitly overridden.
+
 ## Plugins
 
 - Paranoid plugin handles soft delete filtering and must respect aggregate first-stage constraints such as `$geoNear`.
 - Strict reference plugin validates referenced documents and must resolve models through the active connection.
 - Transform plugin centralizes common document transformation and should not break Mongo update operators.
 
+Plugin checklist:
+
+- Paranoid aggregate injection must not come before `$geoNear`.
+- Strict reference checks must be connection-aware in multi-connection apps.
+- Transform behavior must preserve update operators such as `$set`, `$inc`, `$push`, and `$addToSet`.
+- Do not treat plugins as a replacement for app authorization or domain validation.
+
 ## Debug Output
 
 Use `mongoDebug(collection, method, ...args)` when a Mongoose debug callback should be rendered as a Mongo shell-friendly command.
+
+Debug output is for developer diagnostics. Do not log credentials or sensitive document payloads in production logs.

package/dist/codex/skills/joktec-mysql-skill/SKILL.md

@@ -19,8 +19,13 @@ Use this skill for relational resources backed by JokTec's TypeORM wrapper.
 - Treat `mysql`, `mariadb`, and `postgres` as the first-class dialects.
 - Keep `sync` explicit and normally enabled only by an owner process or development bootstrap.
 - Do not add new behavior to deprecated `MysqlFinder`; use `MysqlRepo.qb()` and `MysqlHelper` paths.
+- Use schema-first `@Column`, `@PrimaryColumn`, and `@TimestampColumn` wrappers when an entity also acts as DTO metadata.
+- Use `@Column({ kind: 'virtual' })` for computed getters that need expose/Swagger metadata without persistence.
+- Use `immutable` for API read-only metadata; TypeORM `update: false` remains storage write behavior and is also inferred as Swagger read-only when `immutable` is not set.
+- Do not use `@joktec/mysql` for Mongo/ObjectId columns, even though TypeORM has Mongo-related APIs.
+- For real migrations, inspect the installed `@joktec/mysql` source in the consumer project's `node_modules` first. If that is insufficient, read GitHub package docs, then GitHub source. Use the local `../joktec-framework` checkout only when you are working inside the JokTec development workspace.
 
 ## References
 
 - Read `references/repository.md` for connection lifecycle, repository usage, query safety, transaction, and cursor behavior.
-- Read `references/entities.md` for `@Tables`, `@Column`, `@PrimaryColumn`, uuidv7, and dialect guidance.
+- Read `references/entities.md` for `@Tables`, `@Column`, `@PrimaryColumn`, uuidv7, dialect guidance, and legacy decorator migration rules.

package/dist/codex/skills/joktec-mysql-skill/references/entities.md

@@ -1,14 +1,122 @@
 # MySQL Entity Decorators
 
+## Source Lookup
+
+When blocked in a consumer project, inspect the installed package first:
+
+- `node_modules/@joktec/mysql/README.md`
+- `node_modules/@joktec/mysql/AGENTS.md` when published with the package
+- `node_modules/@joktec/mysql/dist/index.d.ts`
+- `node_modules/@joktec/mysql/dist/decorators/column.decorator.d.ts`
+- `node_modules/@joktec/mysql/dist/decorators/columns/column.type.d.ts`
+- `node_modules/@joktec/mysql/dist/decorators/timestamp.decorator.d.ts`
+
+If the installed package is missing enough detail, use the GitHub package docs next:
+
+- `https://github.com/joktec/joktec-framework/tree/main/packages/databases/mysql`
+
+Use GitHub source only after package docs and installed types are not enough:
+
+- `packages/databases/mysql/src/decorators/table.decorator.ts`
+- `packages/databases/mysql/src/decorators/column.decorator.ts`
+- `packages/databases/mysql/src/decorators/columns/column.type.ts`
+- `packages/databases/mysql/src/decorators/columns/column.factory.ts`
+- `packages/databases/mysql/src/decorators/columns/primary.column.ts`
+- `packages/databases/mysql/src/decorators/columns/timestamp.column.ts`
+- `packages/databases/mysql/src/decorators/columns/virtual.column.ts`
+- `packages/databases/mysql/src/decorators/columns/object.column.ts`
+- `packages/databases/mysql/src/decorators/columns/string.column.ts`
+- `packages/databases/mysql/src/decorators/columns/number.column.ts`
+- `packages/databases/mysql/src/decorators/columns/transform.column.ts`
+- `packages/databases/mysql/src/decorators/columns/swagger.column.ts`
+
 ## Schema-First Entity Pattern
 
-Use `@Tables`, `@Column`, and `@PrimaryColumn` from `@joktec/mysql` when an entity should also act as the source class for mapped DTOs.
+Use `@Tables`, `@Column`, `@PrimaryColumn`, and `@TimestampColumn` from `@joktec/mysql` when an entity should also act as the source class for mapped DTOs.
+
+The new decorators are not thin TypeORM aliases. They are schema-first wrappers that compose:
+
+- TypeORM column/primary column metadata.
+- `class-validator` behavior through `@joktec/utils` validators.
+- `class-transformer` expose/exclude behavior.
+- Swagger property metadata.
+
+The wrapper can also represent virtual computed getters and nested JSON/jsonb class payloads. Do not use this package for Mongo/ObjectId columns.
+
+Wrapper philosophy:
+
+- prefer one schema declaration that carries persistence, validation, transform, and Swagger metadata
+- use wrapper options before duplicating `@ApiProperty`, `@Expose`, `@Type`, or common validators
+- use raw TypeORM only for advanced cases that the wrapper does not model cleanly
+- keep storage write behavior and API documentation behavior distinct when needed
+
+## Current Decorator Capabilities
+
+Use the package README and actual installed source for full migration details. This skill only keeps the common mappings that help an agent recognize old patterns.
+
+Common mappings:
+
+| Legacy decorators | New decorator shape |
+| --- | --- |
+| `@PrimaryGeneratedColumn()` | `@PrimaryColumn('increment')` |
+| `@PrimaryGeneratedColumn('uuid')` | `@PrimaryColumn('uuid')` |
+| app-generated ordered UUID id | `@PrimaryColumn('uuidv7')` |
+| `@CreateDateColumn(...)` | `@TimestampColumn('create', ...)` |
+| `@UpdateDateColumn(...)` | `@TimestampColumn('update', ...)` |
+| `@DeleteDateColumn(...)` | `@TimestampColumn('delete', ...)` |
+| TypeORM `@VersionColumn(...)` | `@Column({ kind: 'version', ... })` |
+| TypeORM `@VirtualColumn(...)` | `@Column({ kind: 'virtual', mode: 'sql', query, ... })` |
+| TypeORM `@ViewColumn(...)` | `@Column({ kind: 'view', ... })` |
+| `@Column(...)` | `@Column(...)` from `@joktec/mysql` |
+| `@RelationId(...)` | `@Column({ kind: 'relation-id', relationId })` |
+| `@IsNotEmpty()` | `@Column({ required: true })` |
+| `@IsOptional()` | `@Column({ required: false })` or nullable TypeORM option when storage allows null |
+| `@IsEmail()` | `@Column({ isEmail: true })` |
+| `@IsMobilePhone()` | `@Column({ isPhone: true })` |
+| `@IsInt()` | `@Column({ isInt: true })` or an integer column type |
+| `@IsUUID()` | `@Column({ isUUID: true })` |
+| `@IsObject()` | `@Column({ isObject: true })` or a JSON column |
+| `@IsHexColor()` | `@Column({ isHexColor: true })` |
+| `@IsUrl()` | `@Column({ isUrl: true })` |
+| `@MinLength(n)` | `@Column({ minLength: n })` |
+| `@MaxLength(n)` | `@Column({ maxLength: n })` |
+| `@Min(n)` | `@Column({ min: n })` |
+| `@Max(n)` | `@Column({ max: n })` |
+| `@Expose()` | default behavior of `@Column(...)` |
+| `@Expose({ groups })` | `@Column({ groups })` |
+| `@Exclude({ toPlainOnly: true })` plus hidden Swagger | `@Column({ hidden: true })` |
+| `@ApiProperty(...)` | `@Column({ swagger: ... })`, or native options such as `example`, `comment`, `deprecated`, `min`, `max`, `minLength`, `maxLength` |
+| `@ValidateNested()` + `@Type(() => Preference)` | `@Column('jsonb', { nested: Preference })` |
+| `@ValidateNested({ each: true })` + `@Type(() => Preference)` | `@Column('jsonb', { nested: Preference, each: true })` |
+| `@Expose()` + `@ApiProperty(...)` on a getter | `@Column({ kind: 'virtual', ... })` |
+| Swagger `readOnly: true` | `@Column({ immutable: true })` or TypeORM `update: false` when ORM updates must also be blocked |
+
+## Read-Only Metadata
+
+`immutable` is the API read-only hint used by the JokTec MySQL wrapper. TypeORM `update: false` is the ORM write behavior. The wrapper maps both to Swagger `readOnly` when appropriate:
+
+- `immutable` has priority over `update: false`
+- `update: false` implies Swagger `readOnly` only when `immutable` is not set
+- `swagger.readOnly` remains the final explicit override
+
+Some field kinds default to API read-only because they are system-managed or computed:
+
+- primary keys
+- timestamp columns
+- version columns
+- view columns
+- virtual getter and SQL virtual columns
+- relation-id columns
+- tree level columns
 
 ## Primary Keys
 
 - Prefer numeric auto-increment keys for write-heavy MySQL tables.
 - Use UUIDs when the app needs globally unique or public identifiers.
 - Prefer `uuidv7` over random UUIDs when the id participates in ordered indexes or cursor-like access.
+- When switching from UUID v4 to uuidv7, verify downstream code does not assume random UUID ordering.
+
+Do not blindly convert every primary key to uuidv7. Keep numeric increment keys when the table is internal, write-heavy, and does not need public/global identifiers.
 
 ## Dialects
 

package/dist/codex/skills/joktec-mysql-skill/references/repository.md

@@ -1,23 +1,86 @@
 # MySQL Repository Usage
 
+## Source Lookup
+
+When blocked in a consumer project, inspect installed package docs and types first:
+
+- `node_modules/@joktec/mysql/README.md`
+- `node_modules/@joktec/mysql/AGENTS.md` when published with the package
+- `node_modules/@joktec/mysql/dist/index.d.ts`
+- `node_modules/@joktec/mysql/dist/mysql.module.d.ts`
+- `node_modules/@joktec/mysql/dist/mysql.service.d.ts`
+- `node_modules/@joktec/mysql/dist/mysql.repo.d.ts`
+- `node_modules/@joktec/mysql/dist/models/mysql.request.d.ts`
+
+If the installed package is insufficient, read GitHub package docs next:
+
+- `https://github.com/joktec/joktec-framework/tree/main/packages/databases/mysql`
+
+Use GitHub source only after installed types and package docs are not enough:
+
+- `packages/databases/mysql/README.md`
+- `packages/databases/mysql/AGENTS.md`
+- `packages/databases/mysql/src/index.ts`
+- `packages/databases/mysql/src/mysql.module.ts`
+- `packages/databases/mysql/src/mysql.service.ts`
+- `packages/databases/mysql/src/mysql.repo.ts`
+- `packages/databases/mysql/src/helpers/mysql.helper.ts`
+- `packages/databases/mysql/src/helpers/mysql.finder.ts`
+- `packages/databases/mysql/src/services/mysql.dialect.ts`
+- `packages/databases/mysql/src/models/*`
+
 ## Module Setup
 
 Register entities with `MysqlModule.forRoot({ conId, models: [...] })`. Use `conId` for multiple DataSources.
 
+Best practice:
+
+- Register consumer app entities in an app repository module.
+- Keep `sync` disabled in request-facing processes unless the app intentionally owns schema sync.
+- Use one controlled owner process for schema sync/migration in multi-process deployments.
+- Preserve `conId` when resolving repositories or transaction-scoped managers.
+
 ## Repository Pattern
 
 Extend `MysqlRepo` and pass the entity class to the base constructor. Services can use `BaseService` when CRUD behavior follows the shared contract.
 
+Repository checklist:
+
+- Keep entity-specific SQL helpers in the app repository, not in controllers.
+- Use `MysqlRepo.qb()` and repository methods for standard reads so field validation, relation loading, soft delete, and pagination stay consistent.
+- Do not add new behavior to `MysqlFinder`; it is deprecated compatibility code.
+- Keep read/write operations in the same transaction context when the app passes a manager or query runner.
+
 ## Query Safety
 
 - Validate field paths against TypeORM metadata before interpolating SQL identifiers.
 - Use parameter binding for values.
 - Keep logical operators such as `$and` and `$or` grouped through QueryBuilder behavior.
 
+Anti-patterns:
+
+- Do not interpolate user-provided field names or relation names into SQL without TypeORM metadata validation.
+- Do not assume MySQL-only syntax when the package supports MySQL, MariaDB, and Postgres.
+- Do not use a relation/populate path unless it maps to entity metadata.
+
 ## Pagination
 
 `MysqlRepo.paginate` supports page, offset, and cursor responses. Cursor mode defaults to `createdAt` plus primary key columns. Custom cursor keys must be mapped columns.
 
+Cursor checklist:
+
+- Use `createdAt + primary key` as the default stable cursor.
+- Add primary keys as tie-breakers for non-unique cursor keys.
+- Ensure cursor keys are indexed for hot list endpoints.
+- Validate cursor payload shape before building SQL.
+- Treat cursor tokens as opaque; clients should not parse them.
+
 ## Transactions
 
 When using transaction-scoped operations, pass the manager or query runner through repository options so pre-read and write operations use the same context.
+
+Transaction checklist:
+
+- Use one manager/query runner for read-modify-write flows.
+- Rollback tests should prove that both pre-read dependent writes and writes are transaction-scoped.
+- Avoid mixing default repositories and transaction-scoped repositories in one operation.

package/dist/codex/skills/joktec-tool-skill/SKILL.md

@@ -23,6 +23,7 @@ Use this skill for reusable utility services.
 - Use config-driven clients instead of direct ad hoc setup in app code.
 - Preserve retry, metrics, proxy, and upload behavior where the package exposes it.
 - Keep alert tokens and webhook URLs in runtime config only.
+- If guidance is insufficient, read this skill's references and inspect `../joktec-framework` package source or GitHub fallback before assuming APIs.
 
 ## Reference
 

package/dist/codex/skills/joktec-tool-skill/references/tools.md

@@ -1,13 +1,49 @@
 # Tool Usage
 
+## Source Lookup
+
+When blocked, inspect:
+
+- `packages/tools/README.md`
+- `packages/tools/AGENTS.md`
+- `packages/tools/<package>/README.md`
+- `packages/tools/<package>/src/index.ts`
+- package config/module/service/helper files under `src/`
+
 ## HTTP
 
 Use `@joktec/http` for Axios-backed requests, uploads, proxy agent support, retry config, and metrics where exposed.
 
+Best practice:
+
+- Use the package service for outbound HTTP so retry/proxy/metrics behavior stays centralized.
+- Keep external endpoint URLs and credentials in runtime config.
+- Be careful with ESM/CommonJS import changes in HTTP/Axios ecosystem packages.
+- `HttpService.buildAgent(proxy, opts)` expects proxy identity in `HttpProxyConfig` and agent tuning in Node `AgentOptions`; inspect the installed source before adapting to proxy-agent major-version changes.
+- Test request behavior with mocks unless the test is an explicit consumer integration scenario.
+
 ## File
 
 Use `@joktec/file` for shared file helpers and classification behavior instead of duplicating local utility code.
 
+Best practice:
+
+- Keep filesystem paths and temporary directories environment-specific.
+- Validate upload/file inputs before passing them into business workflows.
+- Use package helpers for MIME/classification behavior when consistency matters across services.
+
 ## Alert
 
 Use `@joktec/alert` for Slack-compatible webhook alerts. Keep webhook URLs and credentials out of source control.
+
+Best practice:
+
+- Treat alert messages as operational notifications, not business workflows.
+- Do not leak secrets, tokens, connection strings, or personal data into alert payloads.
+- `@joktec/alert` is present but less complete than core database/client packages; inspect source before depending on advanced behavior.
+
+## Anti-Patterns
+
+- Do not scatter raw Axios instances across the app when `@joktec/http` should own shared behavior.
+- Do not commit webhook URLs or proxy credentials.
+- Do not use tool packages as hidden places for app business rules.