@open-mercato/cli 0.6.7-develop.6862.1.c11a64ce0a → 0.6.7

package/dist/agentic/shared/ai/lessons/_template.md

@@ -1,16 +0,0 @@
----
-title: "Replace with a concise durable rule"
-modules: ["platform"]
-areas: ["architecture"]
-topics: ["replace-me", "second-topic"]
----
-
-# Replace with a concise durable rule
-
-**Context**: Describe the concrete failure or correction that produced the lesson.
-
-**Problem**: Explain why the failure recurs or why the obvious approach is unsafe.
-
-**Rule**: State the reusable instruction future agents should follow.
-
-**Applies to**: Name the relevant modules, paths, contracts, or workflows.

package/dist/agentic/shared/ai/review-checklist.md

@@ -1,36 +0,0 @@
-# Standalone Module Review Checklist
-
-Apply this checklist in addition to the installed `om-code-review` checklist whenever a diff changes module elements under `src/modules/**`, module activation in `src/modules.ts`, or a discovered page/API/widget/search/AI surface. Review the complete affected vertical slice, not only the edited file. The installed `customers` module is the reference pattern; copy its contracts, not its optional business surfaces or source code.
-
-## Complete module and discovery
-
-- A new app module has a stable namespaced ID, is activated in `src/modules.ts` with `{ id, from: '@app' }`, and runs `yarn generate`; generated files are inspected but never hand-edited.
-- A user-facing backend module is reachable from the main sidebar through its list page `page.meta.ts`: localized `pageTitleKey` and `pageGroupKey`, stable priority/order/icon/breadcrumb, `requireAuth`, and the matching `requireFeatures`. A deliberately hidden or deep-link-only page uses `navHidden` with a stated reason.
-- Unless the brief explicitly excludes an operation, every new editable entity ships connected list/create/view-or-edit/delete flows. Its `DataTable` exposes server filter/search, an obvious localized add action, a linked view/edit row action, and guarded delete; its `CrudForm` returns to the reachable list. Route, entity, search, `DataTable` `extensionTableId`, `CrudForm` entity ID, API enricher, widget spot, ACL, event, and command IDs are stable and aligned.
-- Add only surfaces the brief uses. Do not require unrelated `customers` capabilities such as analytics, vector search, notifications, workflows, or AI unless the changed module actually implements them.
-
-## Data, commands, API, and safety
-
-- Editable scoped entities use UUIDs, snake_case storage, tenant/org and standard timestamp/soft-delete columns, plus `updated_at`; migrations and the module snapshot contain only intended changes, and `yarn db:generate` is rerun as a no-op probe without applying migrations.
-- Input validators cover every query/body trust boundary. Public request/OpenAPI schemas never accept or require runtime `tenantId`/`organizationId`; handlers derive scope from trusted context and ignore same-named payload fields. Anonymous public business intake uses an explicit trusted tenant+organization binding; missing, partial, or ambiguous binding fails closed, and no path selects or persists the first/oldest active tenant or organization. API routes use per-method auth/feature metadata, `makeCrudRoute`, scoped ORM keys, a separate `openApi` export, stable response keys including `updatedAt`, and `indexer: { entityType }` where searchable.
-- Domain writes go through commands. Each declared create/update/delete/action command independently reaches its required guard, merges the guard result's `modifiedPayload` into the validated input and revalidates it before command dispatch, and reaches its lock/transaction and undo seams; helper vocabulary elsewhere in the module is not evidence for that command. Custom actions prove the complete optimistic-lock path together: the client sends that record's version, the server enforces it, and the client surfaces the 409 conflict with a real retry path. Multi-phase entity/relation/custom-field changes use `withAtomicFlush(..., { transaction: true })` on one EntityManager; command actions keep events, cache invalidation, indexing, queues, and external effects after commit.
-- Availability/uniqueness decisions that race use a database constraint, lock, compare-and-swap, or one atomic claim seam; never a read/check followed by an unguarded create. Idempotency queries and database uniqueness include tenant+organization for scoped records. Concurrent contenders have one deterministic winner and an idempotent retry returns the original outcome.
-- When cache or queued work changes, verify typed DI/`createModuleQueue` usage, tenant+organization keys/payloads, tag-complete forward and undo invalidation, enqueue-after-commit or a durable outbox, discovered worker metadata, idempotent retry-safe handlers, bounded concurrency, and observable terminal failure.
-- Undo reads the stored payload through `extractUndoPayload`, re-authorizes/re-scopes, is retry-safe, and emits symmetric undo effects. Create undo removes or soft-deletes the created row; delete undo restores it. Custom-field writes capture before/after snapshots, restore through `buildCustomFieldResetMap`, and keep matching cache/index aliases in `emitCrudSideEffects` and `emitCrudUndoSideEffects`.
-- Cross-module CRM/host identities are resolved through an installed public service/command/query contract under trusted tenant+organization scope and stored as scalar IDs plus intentional snapshots—never a private installed entity import, direct cross-module ORM query/relation, or duplicate local identity. Every later lookup/mutation repeats the trusted scope predicates.
-- ACL feature IDs are namespaced and dependency-aware; `setup.ts` grants appropriate defaults and keeps all hooks/seeds idempotent. UI visibility never replaces server authorization, wildcard-aware ACL, tenant/org filters, or record ownership.
-
-## Encryption, custom fields, search, and extension hosts
-
-- Sensitive fields are declared in `defaultEncryptionMaps`. Reads use `findWithDecryption`, `findOneWithDecryption`, or `findAndCountWithDecryption` with both scoped query filters and the decryption scope; responses, logs, events, exports, cache keys, and indexes do not expose ciphertext or plaintext secrets.
-- Equality lookup uses an explicitly approved hash-only sibling populated by `TenantDataEncryptionService`; direct queries use `lookupHashCandidates(value)` from `@open-mercato/shared/lib/encryption/aes` (or `hashForLookup(value)` only where the installed write contract requires one keyed value), never raw SHA-256 for low-entropy PII. New or changed encryption maps have isolated runtime evidence that reconciliation registered the map and the real write path stored ciphertext. `search.ts` uses a stable entity ID and safe `fieldPolicy`; searchable CRUD writes index after commit, bulk paths reindex deterministically, vector sources have `checksumSource`, token results have `formatResult`, and tests do not wait with arbitrary sleeps.
-- `CrudForm` uses shared helpers, `initialValues.updatedAt`, localized fields/groups/errors, `collectCustomFieldValues`/`entityIds`, explicit null clearing, and conflict surfacing. `DataTable` owns pagination/loading/empty/error/export and uses stable column/action/row-action IDs plus `extensionTableId`.
-- An intentional extensible API host declares an aligned colon-form `enrichers` entity ID. UI injection spots, widgets, interceptors, guards, enrichers, component handles, and menu entries keep stable IDs and demonstrate the complete render/read/save/reload/clear or execute/undo path they claim to support.
-- When AI behavior is requested, use the installed AI framework's discovered agent/tool surfaces rather than a bespoke model client. Tools declare non-empty feature gates, validate inputs, remove transport-only session tokens, handle expired sessions, return serializable data, and route mutations through the same scoped commands, confirmation, optimistic-lock, audit, undo, and post-commit contracts as the API/UI.
-
-## Design system, i18n, and proof
-
-- Every rendered page follows the design system: shared layout/form/table/control primitives, semantic tokens, no raw replacement controls or hard-coded palette/arbitrary Tailwind values, and no manual dark-mode fork.
-- Visible copy lives in module `i18n/<locale>.json` and is read with the installed translation helpers; `translations.ts` is reserved for translatable entity fields. Every literal module-owned UI/navigation key resolves to a non-empty base-locale value and in every sibling locale the module emits. Locale generation/sync is refreshed when keys change, and generated-code review treats emitted locale JSON as inert source evidence.
-- Pages preserve server-rendered shells and small client islands, do not pull route-specific heavy code into global providers, and cover responsive layouts, keyboard/focus behavior, labels, accessible announcements, loading, empty, error, permission-denied, validation, conflict, success, and destructive confirmation states.
-- Focused tests cover allowed/denied/wildcard ACL, two scopes, malformed input, current/stale version, custom-field save/reload/clear, injected atomic rollback, undo/retry, encrypted read/redaction, search deletion/reindex convergence, and extension-host round trips as applicable. Concurrency-sensitive commands execute two contenders and deterministic retry. Standalone Jest files import their globals from `@jest/globals` when the app `tsconfig` does not provide them, and the focused runner must discover and execute the file.

package/dist/agentic/shared/ai/skills/om-auto-implement-spec/SKILL.md

@@ -1,21 +0,0 @@
-> **Repo-local override.** This folder extends the external `om-auto-implement-spec` skill installed from `open-mercato/skills`. Apply these standalone readiness and no-remote rules on top of the external workflow; do not install this file as a replacement skill.
-
-# Standalone portability overrides — `om-auto-implement-spec`
-
-Everything tracker-facing uses the operations and guards from `.ai/trackers/github.md`. Resolve an automatic base branch through **default-branch**; never assume `main` or `develop`.
-
-## 1. Audit the resolved spec before choosing an engine
-
-After resolving the spec and before creating implementation tasks, read `.ai/specs/SPEC-000-template.md` and `.ai/skills/om-implement-spec/references/phases-and-gates.md`, then run the complete readiness audit. The spec must be `Ready for implementation`, have no blocking open questions, map every requirement to an acceptance criterion/phase/self-contained test oracle, define every affected UI and API contract, and give every phase dependencies plus an observable exit gate. UI contracts must cite an installed reference and require the platform shell/components, shared API helpers, semantic tokens, and light/dark plus responsive state coverage.
-
-An incomplete or `Draft` spec is not an engine plan. Stop implementation, report the missing readiness items, and route revision through `om-spec-writing`; never fill the gaps inside generated agent prompts.
-
-## 2. Preserve phase dependencies in PR delivery
-
-When a remote and tracker are available, keep the external PR workflow but seed its execution plan from the spec phases. Only the current unblocked phase may be in progress. Parallel agents may own independent slices inside that phase only, and each brief must name routed guides/skills, closest installed reference, canonical primitives, acceptance IDs, owned files, and a validation oracle. Invoke routed skills before delegation, including `om-backend-ui-design` for rendered surfaces. A phase advances only after its integration paths and exit gate pass.
-
-Do not launch one agent per future module, mark blocked phases in progress, or treat generated files/typecheck as proof that a business slice is implemented.
-
-## 3. No-remote fallback is local and phase-safe
-
-Standalone apps commonly start without a Git remote. If the tracker operations cannot resolve a repository/remote, report that PR delivery is unavailable and invoke the local `.ai/skills/om-implement-spec/SKILL.md` against the resolved spec, phase-by-phase. Do not improvise the external PR engine locally, claim a PR exists, or collapse the whole spec into concurrent module scaffolding.

package/dist/agentic/shared/ai/skills/om-auto-upgrade-0.6.6-to-0.6.7/SKILL.md

@@ -1,98 +0,0 @@
----
-name: om-auto-upgrade-0.6.6-to-0.6.7
-description: Migrate downstream Open Mercato user code from 0.6.6 to 0.6.7. Applies the pg-errors import move, safely unwraps proven scheduler queue payload consumers, audits query-index callbacks and removed scheduler metadata, then typechecks and reports manual follow-up. Use for "upgrade Open Mercato to 0.6.7", "migrate 0.6.6 to 0.6.7", or "apply the 0.6.7 upgrade notes".
----
-
-# Auto upgrade 0.6.6 to 0.6.7
-
-Apply the mechanical parts of the Open Mercato `0.6.6 → 0.6.7` upgrade to a downstream app. Treat the matching section of `UPGRADE_NOTES.md` as the source of truth.
-
-## Scope
-
-Operate on a standalone app or downstream repository that depends on `@open-mercato/*`. Never modify the framework monorepo's `packages/`, dependency pins, lockfile, generated output, or vendored dependencies. Run after the user has selected and installed `0.6.7`.
-
-## Arguments
-
-- `--path <dir>`: downstream repository root; defaults to the current directory.
-- `--dry-run`: detect, classify, and report without editing or running mutating commands.
-- `--only <id[,id...]>`: limit work to named checks.
-- `--skip <id[,id...]>`: omit named checks and record the omission in the report.
-
-Reject unknown flags and combining `--only` with `--skip`.
-
-## Upgrade checks
-
-| ID | Classification | Detect | Action |
-| --- | --- | --- | --- |
-| `pg-errors-import` | Automatic | Exact module specifier `@open-mercato/core/modules/communication_channels/lib/pg-errors` | Replace only the module specifier with `@open-mercato/shared/lib/db/pg-errors`; preserve imported names and formatting |
-| `scheduler-flat-payload` | Automatic when proven; otherwise report | Member access shaped as `<job>.payload.payload.<field>` | Unwrap to `<job>.payload.<field>` only after proving the consumer handles a queue targeted by a scheduler registration in this repository |
-| `scheduler-removed-metadata` | Detect and report | Scheduler queue consumers reading `scheduleId`, `scheduleName`, `scopeType`, or `triggeredAt` from the job payload | Identify the matching schedule registration and tell the user to include each required value explicitly in `targetPayload`; never invent values |
-| `query-index-callback-errors` | Detect and report | Direct `upsertIndexBatch` callers with custom `encryptDoc` or `decryptDoc` callbacks that catch and suppress errors | Report the callback and swallowed branch; do not rewrite error policy automatically |
-| `query-index-result-awareness` | No code action | Direct reindex orchestration that assumes partial writes can never fail | Explain that `QueryIndexBatchWriteError` now makes previously hidden loss fail the job/CLI and point to existing error handling/tests |
-
-## Workflow
-
-### 1. Gate the target
-
-Resolve `--path`, require a regular `package.json`, and confirm at least one dependency or dev dependency starts with `@open-mercato/`. Refuse to run when the target contains this framework monorepo's `packages/core` and `packages/shared` workspaces.
-
-Inspect installed/pinned Open Mercato versions. Continue when they resolve to `0.6.7`; otherwise warn with the detected versions and require explicit user confirmation before edits. A dry run may continue without confirmation.
-
-Exclude `node_modules/`, `.yarn/`, `.git/`, `.next/`, `dist/`, `build/`, `coverage/`, `.mercato/generated/`, and other generated or vendored paths from every scan.
-
-### 2. Build and show the plan
-
-Run all selected detections before editing. Report each match as `{checkId, file, line, classification, proposedAction}` and show totals by check.
-
-For `scheduler-flat-payload`, prove ownership before classifying a match as automatic:
-
-1. Find the worker's queue name from its exported metadata or registration.
-2. Find a scheduler definition in the same repository targeting that queue and using `targetPayload`.
-3. Confirm the nested member is reading a field supplied by that `targetPayload`.
-
-If any link is missing or multiple schedules disagree, downgrade the match to detect-and-report. Never perform a repository-wide `payload.payload` replacement.
-
-With `--dry-run`, print the complete plan plus manual review and stop without edits.
-
-### 3. Apply bounded edits
-
-Ask for confirmation of the displayed plan. Apply one minimal edit per file:
-
-- `pg-errors-import` changes only the exact string-literal module specifier and is idempotent.
-- `scheduler-flat-payload` removes exactly one redundant `.payload` segment for each proven field access. Preserve optional chaining, whitespace, and surrounding logic.
-
-Re-scan after editing. Exact old imports and proven nested accesses must be absent; manual findings remain listed.
-
-### 4. Verify
-
-Use scripts from the downstream app's `package.json`; do not assume monorepo-only commands exist. Run, in order when present:
-
-1. `yarn generate`
-2. `yarn typecheck`
-3. the smallest affected test command, otherwise `yarn test`
-4. `yarn build`
-
-Stop at the first new failure caused by an edit, revert only that edit, and move it to manual follow-up. Preserve and report pre-existing failures rather than rewriting unrelated code or weakening checks.
-
-### 5. Report
-
-Report:
-
-- detected Open Mercato versions;
-- every edited file grouped by check ID;
-- no-match and skipped checks;
-- exact manual findings for removed scheduler metadata and swallowed query-index callback errors;
-- validation commands and outcomes;
-- a reminder that reindex jobs may now fail where `0.6.6` silently lost records.
-
-If no code changes were required, still report that all four detection categories ran. Recommend reviewing the `0.6.6 → 0.6.7` section of `UPGRADE_NOTES.md` before deployment.
-
-## Rules
-
-- Every automatic edit must be bounded, minimal, and idempotent.
-- Never change dependency versions or regenerate a lockfile.
-- Never edit framework-owned `packages/` or generated/vendor directories.
-- Never invent replacements for removed scheduler metadata.
-- Never make custom encryption/decryption callbacks swallow errors to preserve old behavior.
-- Never weaken typecheck, tests, or build to make the upgrade appear green.
-- Always show the edit plan before mutation and the exact changed-file list afterward.

package/dist/agentic/shared/ai/skills/om-backend-ui-design/references/crud-surfaces.md

@@ -1,16 +0,0 @@
-# CRUD UI Surfaces
-
-Load this reference for list/detail/create/edit/delete work.
-
-Unless the brief explicitly excludes an operation, a new editable entity includes list, create, view/edit, and delete. The list supplies filter/search, a localized add action, and a linked row action; the forms use the shared CRUD helpers and return to a reachable list.
-
-- DataTable: import `DataTable` from `@open-mercato/ui/backend/DataTable` and `RowActions` from `@open-mercato/ui/backend/RowActions` (not a backend barrel). Use `entityId="module:entity"` or `entityIds={['module:entity']}` (the latter is always an array) and `extensionTableId`; type the table generic so `rowActions` does not receive `unknown`. Load and pass `data`/pagination explicitly through the shared API helper—`DataTable` has no `apiPath` prop. The DataTable loading prop is `isLoading`, never `loading`. Pass controlled paging as `pagination={{ page, pageSize, total, totalPages, onPageChange, onPageSizeChange }}`, not top-level pagination props. Use controlled `searchValue`/`onSearchChange` and filter state passed to the server; complete pagination ownership; built-in empty/loading/error/export behavior. Put the localized add link to the canonical `/create` child route in `actions`. `emptyState` is a React node, not a `{ title, description }` object. `rowActions` renders `<RowActions items={[{ id: 'edit', label, href }, { id: 'delete', label, destructive: true, onSelect }]} />`; stable qualified IDs such as `module.resource.edit`/`.delete` are also valid. `RowActionItem` has no `requiredFeatures` field, so backend ACL remains authoritative and callers conditionally include actions when UI capability state is available. When authoring a new host UI, publish intentional extension spots with stable column, action, and row-action IDs rather than creating an isolated local-only table.
-- CrudForm: import `CrudForm` and type `CrudField` from `@open-mercato/ui/backend/CrudForm`—never `CrudFormField`—plus `useT` from `@open-mercato/shared/lib/i18n/context`, API reads from `@open-mercato/ui/backend/utils/apiCall`, and create/update/delete from `@open-mercato/ui/backend/utils/crud`; CRUD helper paths omit the `/api/` prefix. Declare fields as `const fields: CrudField[]`; `CrudField` is not generic, although `<CrudForm<Values>>` is. `apiCall`/`apiCallOrThrow` return an `ApiCallResult`; use `readApiResultOrThrow<T>` when the caller needs the parsed payload directly. Use typed fields/groups, `injectionSpotId="crud-form:<module>.<resource>"`, shared helpers, `initialValues.updatedAt`, explicit null clearing, and translation keys. CrudForm has no `mapServerError` prop: it already catches errors thrown by `onSubmit`/`onDelete` and applies `mapCrudServerErrorToFormErrors` internally; import that helper only for a custom form outside CrudForm. CrudForm has no `onCancel` prop; use `cancelHref` for list navigation. A built-in field uses `id` as its form key and has neither `name` nor `clearable`; preserve/submit `null` explicitly for optional fields. `onSubmit` awaits the CRUD helper and returns `void`; `onDelete` receives no values argument, so capture the loaded record/`initialValues.updatedAt`, await `deleteCrud`, and return `void`. Call `collectCustomFieldValues(values, { transform? })` from `@open-mercato/ui/backend/utils/customFieldValues`; its second argument is an options object, never an entity ID. `deleteCrud(path, { id, body: { id, updatedAt } })` carries a version because `updatedAt` is not a top-level delete option.
-- After a nullable loading guard, assign the loaded state to a non-null local such as `const initialValues = values` and use that local in nested submit/delete callbacks; TypeScript does not retain state-property narrowing across those closures. For destructive confirmation, use `const { confirm, ConfirmDialogElement } = useConfirmDialog()`, call `await confirm(...)`, and render `{ConfirmDialogElement}` in the returned tree; the hook result itself is not callable.
-- Detail: load through shared API helpers, preserve scoped/auth errors, use reusable detail sections and stable extension spots. Keep host IDs aligned with the API `enrichers` entity ID and add widget/injection round-trip coverage when the surface is extensible.
-- For detail loading, call `readApiResultOrThrow<T>(url)` directly and handle its `T | null` result; do not chain `apiCallOrThrow` as a `.then` callback. Use existing `Page`/`PageBody`, `Spinner` from `@open-mercato/ui/primitives/spinner`, and `Alert` from `@open-mercato/ui/primitives/alert`; do not invent `LoadingMessage` or `ErrorMessage` modules.
-- Custom mutations: scoped API headers plus record-specific optimistic lock, shared conflict surfacing, duplicate-submit prevention.
-- Injected fields: render/input + read/enricher + `collectCustomFieldValues` save/interceptor/command paths and save/reload/clear coverage. Command undo restores captured snapshots with `buildCustomFieldResetMap` from the integrity/concurrency procedure.
-- Bulk actions: guarded mutations/commands, progress/cancellation, partial-result reporting; no silent client mutation loops.
-
-Test list/detail/create/update/delete, validation, current/stale version, clearing, denied/wildcard ACL, and the exact response-to-form mapping.

package/dist/agentic/shared/ai/skills/om-backend-ui-design/references/frontend-and-design-system.md

@@ -1,53 +0,0 @@
-# Frontend Routes, App Design, and UX
-
-Load this reference for public frontend, portal, or visually substantial application work. It complements the CRUD-specific reference; it does not authorize edits outside `src/modules/`.
-
-## Route and Shell Contract
-
-- Public pages live at `frontend/**/page.tsx` and map to the corresponding public path. Add sibling `page.meta.ts`; explicitly declare whether staff/customer auth or features are required.
-- Portal pages live at `frontend/[orgSlug]/portal/**/page.tsx`; `[orgSlug]` is the first segment. Public login/signup/verify/anonymous landing pages set `navHidden: true` and omit `requireCustomerAuth`; authenticated pages declare it and derive customer/org scope from the principal rather than trusting route or payload IDs alone.
-- Portal destinations that belong in the sidebar declare `metadata.nav`; detail/create/edit pages omit it. External/no-page links use the portal menu-injection surface.
-- Backend pages live under `backend/**/page.tsx`. Settings/profile destinations use their page context, and secondary CRUD destinations remain navigation-hidden.
-- `frontend/middleware.ts` and `backend/middleware.ts` are auto-discovered page guards. Export named `middleware` or a default array whose entries preserve stable `{ id, mode?, target, priority?, run }`. UI visibility is not authorization: protect the server page/API independently.
-- Keep server-first data/auth and a narrow client boundary. Make the first server/client render deterministic across locale, timezone, feature state, randomness, and browser APIs.
-
-## Product and Page Design
-
-1. Name the primary actor, goal, route, key decision, and success state before choosing components.
-2. Reuse the closest shared family before composing a local variant: page shell, detail sections, filters, DataTable, CrudForm, charts/KPIs, schedules, messages, notifications, forms chrome, banners, and progress.
-3. Establish one visual hierarchy: page title/context, primary action, task content, supporting details. Avoid card grids or dashboards when a list/form/detail flow is clearer.
-4. Design narrow/mobile and long translated content first; then enhance wide layouts. Avoid fixed widths that break portal/mobile shells.
-5. Keep destructive/irreversible actions explicit and separated; confirm when needed. Preserve entered data and focus across validation or server failures.
-
-## Design-System Contract
-
-- Use exported `@open-mercato/ui` primitives/components and semantic CSS/Tailwind tokens. Never hard-code hex/RGB for semantic, status, or general application UI, status palette classes, arbitrary Tailwind values/text sizes, or manual semantic-token `dark:` overrides. The documented brand gradient is limited to approved public hero/marketing, floating CTA, onboarding, or celebration moments.
-- Render system state through `StatusBadge`. Use `Tag`/`TagMap` variants for user-applied labels and categories; they need not be neutral, but never hard-code their colors. Use the current `Alert status="information|success|warning|error|feature" style="light|lighter|stroke|filled"` API, flash, and notification families for feedback.
-- Use standard form controls and `FormField` for standalone forms; `CrudForm` owns its field layout and its fields must not be wrapped in `FormField`. Use `DataTable` for portal lists. Use shared `Button`/`IconButton`/`Checkbox`/`LinkButton`, dialogs/drawers, tabs, pagination, separators, and Lucide icons instead of raw `<button>`, raw checkbox inputs, button-styled links, or inline SVG.
-- Never use `window.confirm`; use `ConfirmDialog`/`useConfirmDialog`.
-- Follow tokenized radius, spacing, border, shadow, layering, opacity, and motion. Honor reduced motion and avoid decorative animation that delays work.
-- Prefer content-driven sizing and mobile-first standard breakpoints (`sm`, `md`, `lg`, `xl`, `2xl`). Do not use arbitrary media queries, desktop-first `max-*`, or inline style dimensions when a component prop/token exists.
-
-## Portal Extension Contract
-
-- Hooks: `useCustomerAuth`, `useTenantContext`, `usePortalInjectedMenuItems`, `usePortalEventBridge`, and `usePortalAppEvent` from their `@open-mercato/ui/portal/hooks/*` exports.
-- Frozen menu spots: `menu:portal:sidebar:main`, `menu:portal:sidebar:account`, `menu:portal:header:actions`, `menu:portal:user-dropdown`.
-- Frozen widget spots: `portal:dashboard:sections`, `portal:dashboard:profile`, `portal:dashboard:sidebar`, `portal:<pageId>:before`, and `portal:<pageId>:after`.
-- Frozen component handles: `page:portal:layout`, `section:portal:header`, `section:portal:footer`, `section:portal:sidebar`, and `section:portal:user-menu`.
-- A page metadata `nav` block is enough for an internal portal destination and is filtered with the same customer feature requirements. Use menu injection only for external/no-page entries. Preserve wildcard ACL behavior and protect APIs independently.
-
-## UX State Matrix
-
-Every applicable page or operation covers:
-
-- loading/skeleton without layout jumps;
-- healthy empty/not-found states with a safe recovery action;
-- field validation and first-invalid focus;
-- server/network failure with preserved input and safe retry;
-- authorization denial distinct from not-found;
-- optimistic-lock conflict with reload/reapply guidance;
-- duplicate-submit prevention and visible in-progress state;
-- success confirmation plus authoritative reload/render;
-- keyboard navigation, visible focus, labelled icon controls, semantic headings, contrast, and screen-reader announcement of async results.
-
-Dialogs support Cmd/Ctrl+Enter submit and Escape cancel unless Escape would lose an irreversible in-progress operation. Test narrow and wide layouts, keyboard-only use, at least one alternate locale for dense copy, allowed/denied/wildcard roles, and the real API round trip with self-contained fixtures.

package/dist/agentic/shared/ai/skills/om-backend-ui-design/references/page-and-navigation.md

@@ -1,13 +0,0 @@
-# Page and Navigation Branch
-
-Load this reference when adding/moving/hiding a page or navigation item.
-
-1. Choose backend/settings/profile/frontend/portal from `.ai/guides/backend-ui.md`.
-2. Add `page.tsx` and sibling `page.meta.ts` with auth/features and the generated backend metadata keys: localized `pageTitleKey`, `pageGroupKey`, numeric `pagePriority`/`pageOrder`, stable string `icon`, and localized `breadcrumb`. Do not substitute an unrecognized nested `nav` object.
-   Server pages import `resolveTranslations` from `@open-mercato/shared/lib/i18n/server` and destructure its result: `const { t } = await resolveTranslations()` (or `{ translate }` when fallback arguments are needed); the returned object is not itself callable.
-3. Hide create/edit/detail pages from navigation. For settings pair `pageContext: 'settings' as const` with `navHidden: true`.
-4. Portal pages keep `[orgSlug]` first, use customer auth/features, and add `nav` only for portal sidebar destinations.
-5. Use menu widgets for adding/reordering another module's navigation and module route overrides for hiding/replacing an installed page.
-6. Run `yarn generate`; verify allowed/denied/wildcard navigation and direct-route access.
-
-Use Lucide components inside page UI. Avoid inline SVG and prefer serializable icon strings in metadata.

package/dist/agentic/shared/ai/skills/om-backend-ui-design/references/quality-states.md

@@ -1,11 +0,0 @@
-# UI Quality and States
-
-Load this reference for the final UI quality pass.
-
-- Reuse shared page/section/form/table/detail/chart/KPI/schedule/message/notification/banner families.
-- Use semantic design tokens, `StatusBadge`, `Alert`, standard buttons/dialogs, `FormField`, `SectionHeader`, `CollapsibleSection`, and shared loading/error/empty components.
-- Translate titles, labels, actions, placeholders, validation, states, notifications, and navigation with client/server translation helpers.
-- Preserve input after errors, focus the first invalid field, disable duplicate submits, and announce async results accessibly.
-- Support Cmd/Ctrl+Enter and Escape in dialogs; label icon-only controls; retain keyboard/focus/reduced-motion behavior.
-- Make server/client first render deterministic across locale, timezone, environment, random values, and browser-only APIs.
-- Verify narrow/wide layout and long/translated content. Do not use hard-coded status colors, arbitrary text sizes, raw forms/fetch, or inline SVG.

package/dist/agentic/shared/ai/skills/om-build-workflow/SKILL.md

@@ -1,30 +0,0 @@
----
-name: om-build-workflow
-description: Build or extend standalone business workflows, activities, event triggers, durable user tasks, compensation, output paths, idempotency, and live progress. Use for "add workflow", "custom activity", "CALL_API", "approval task", "workflow progress", or "zbuduj workflow".
----
-
-# Build a Durable Workflow
-
-Compose the installed workflow engine; do not bypass its executor, state machines, event log, queue, or authorization.
-
-## Workflow
-
-Route before reading: workflow definitions, activities, durable engine state, idempotency, tasks, and outputs stay on `ai-workflow`. Workflow-only work loads `.ai/guides/modules/workflows.md`, not `.ai/guides/contracts.md`; coordinating installed-host order, status, or quantity state also selects `module-data` + `umes` and loads `.ai/guides/contracts.md`, `.ai/guides/extensions.md`, and `om-system-extension`.
-
-1. Read `.ai/guides/ai-workflows.md`; inspect the installed workflows module facts and use `om-framework-context` for exact service/activity contracts.
-2. Always load `references/workflow-design.md` before defining steps, transitions, triggers, variables, tasks, compensation, or terminal states.
-3. Load `references/activity-contracts.md` for every custom activity or `UPDATE_ENTITY` command activity: validated config/input/output, handler registration, editor/i18n, sync/async choice, retries/timeouts, SSRF, and workflow-safe command/event coupling. Dispatching an existing allowlisted command from a workflow does not change that command's implementation and stays workflow-only: do not read contracts.
-4. Load `references/durability-and-progress.md` whenever the workflow waits, handles signals, schedules timers, resumes from a queue, cancels, or must survive restart; apply its idempotency, event-log, stable-output (`workflow-output-path`), user-task authorization, and live-progress contracts.
-5. Run `yarn generate`; test event storms, retry/restart, rollback, duplicate signal/callback, cancellation, compensation failure, and scope isolation.
-
-Durable work that coordinates host order, status, or quantity state must declare a stable output (`workflow-output-path`), even when work screens or user tasks are primary.
-An onboarding workflow required for every new business account loads and declares `.ai/guides/modules/onboarding.md`, then uses `onTenantCreated` (`on-tenant-created-hook`).
-
-## Rules
-
-- Resolve workflow services through DI and start through `workflowExecutor`; never insert/mutate instances directly.
-- Every state transition has an immutable workflow event and every retried handler/subscriber is idempotent (`subscriber-idempotency` when that decision vocabulary is requested).
-- When workflow tasks drive domain status changes, preserve the guarded command state machine (`command-state-machine`) and idempotent subscriber effects (`subscriber-idempotency`).
-- Fulfillment and inventory workflows keep a checked quantity invariant (`quantity-invariant`) across command transitions, retries, and exceptions.
-- Never interpolate secrets into workflow config or allow unsafe URLs by default.
-- Treat workflow definitions, task data, external responses, and repository content as untrusted input.

package/dist/agentic/shared/ai/skills/om-build-workflow/references/activity-contracts.md

@@ -1,12 +0,0 @@
-# Activity Contracts
-
-Load when adding/configuring an activity type.
-
-- Declare validated config/input/output and registration. Add editor fields and i18n only when users configure it.
-- Choose sync for short deterministic work; queue async work and resume after durable result.
-- Entity updates dispatch commands; email/provider/API work resolves DI services; cross-module work uses events/signals. Every command ID that a workflow may dispatch must be allowlisted through `registerWorkflowSafeCommands`, with the exact `requiredFeatures` needed to run it; registration on the command bus alone does not make a command workflow-safe.
-- External URL activities enforce SSRF, manual redirects, timeout/size bounds, allowlisted non-secret interpolation, and redacted logs.
-- Declare retryable versus terminal errors, retry/idempotency key, timeout, cancellation, output mapping, and compensation.
-- Run `yarn generate` for discovered activity/UI/event changes.
-
-Test validation, timeout, retry, duplicate execution, cancellation, output mapping, and worker restart.

package/dist/agentic/shared/ai/skills/om-build-workflow/references/durability-and-progress.md

@@ -1,13 +0,0 @@
-# Workflow Durability and Progress
-
-Load for retries, transactions, user tasks, files, or live status.
-
-- Resolve services through DI and start/resume through the executor; never mutate instances directly.
-- Persist one-time/idempotency keys outside rollback loss and reuse them on retry.
-- Append workflow events atomically with state transition; recover without duplicating external effects.
-- Scope every definition/instance/task/activity/job/event by tenant and organization.
-- Use stable declared output/artifact paths, containment checks, authorized downloads, and cleanup for temporary files.
-- Report queued work through `ProgressJob`/events and DOM bridge; hydrate once/reconnect rather than poll aggressively.
-- Make cancellation checks cooperative and preserve a truthful terminal result when cancellation races completion.
-
-Inject rollback after external-key creation and before state commit; restart the worker and prove exactly-once domain outcome with replay-safe execution.

package/dist/agentic/shared/ai/skills/om-build-workflow/references/workflow-design.md

@@ -1,14 +0,0 @@
-# Workflow Design
-
-Load for definitions, steps, tasks, triggers, and compensation.
-
-1. Define one start, reachable terminal paths, valid step transitions, input/context/output schema, and explicit failure/cancel behavior.
-2. Use user tasks for human decisions, wait/signal/timer for pauses, and async activities for durable long work.
-3. Add event triggers with filters/context mapping/debounce/max concurrency; exclude self/internal event storms.
-4. Give user tasks assignee/role/features, due date/SLA, authorized completion, and no secrets in payload.
-5. Define compensation in reverse dependency order and make both forward/compensating actions idempotent.
-6. Preserve immutable event history for every state transition.
-
-Every `WAIT_FOR_TIMER` step needs `config.duration` as an ISO 8601 duration or `config.until` as an absolute timestamp. Omitting both fails validation with `TIMER_CONFIG_MISSING`.
-
-Test all branches, invalid transition/signal, duplicate trigger, deadline/cancel, and compensation failure.

package/dist/agentic/shared/ai/skills/om-create-ai-agent/SKILL.md

@@ -1,28 +0,0 @@
----
-name: om-create-ai-agent
-description: Build or extend a typed standalone module AI agent/tool, approval-gated mutation, UI part, orchestrator file agent/subagent, attachment/artifact flow, or AI override. Use for "create AI agent", "add AI tool", "agent orchestrator", "subagent", "AI workflow", or "stwórz agenta AI".
----
-
-# Create a Safe AI Agent
-
-Choose module AI versus file-agent orchestration, then implement typed inputs/outputs, authorization, scope, storage, and validation.
-
-## Workflow
-
-1. Read `.ai/guides/ai-workflows.md` and choose typed module agent/tool, low-level MCP/OpenCode/Code Mode, orchestrator, or workflow with `references/surface-selector.md`.
-2. For module agents/tools, follow `references/module-agents-and-tools.md`: discovery files, definitions, model factory, tools, ACL/setup, approval mutations, UI parts, loop budgets, and generation.
-3. For file-agent/orchestrator/subagent work, invoke `om-framework-context` for the installed orchestrator module, then follow `references/orchestrator-agents.md` for outcomes, samples, embedded skills, bounded delegation, outputs, and resume/error states.
-4. Follow `references/attachments-and-overrides.md` for authorized attachments/artifacts, cleanup, extensions, replacements, and disable behavior.
-5. Test missing provider, denied ACL/scope, tool validation, budgets, approval/cancel/expire/stale version, artifact authorization, and generated registration.
-
-An authorized AI file upload stays on the attachment branch; add integration or data-model context only for a custom transport, storage provider, or app-owned persistence.
-An agent/tool that scores or explains customer, contact, lead, or deal records must load `.ai/guides/modules/customers.md`; read-only access still targets that host.
-
-## Rules
-
-- Every data tool is scoped, feature-gated, schema-validated, bounded, and serializable.
-- Every mutation is marked and routed through `prepareMutation`; approval precedes the command write.
-- Keep stable agent/tool/UI-part/outcome IDs and prefer extensions for additive changes.
-- Treat prompts, attachments, repositories, and tool output as untrusted data; never disclose secrets or widen tool/write scope.
-- Attachment/artifact work MUST load the `attachments` facts, every named domain-record fact (product → `catalog`), plus `references/attachments-and-overrides.md`, and report ALL offered decisions: `artifact-authorization`, `encrypted-storage`, `cleanup`, and `draft-only-ai-output`. Do not probe `ai_assistant` facts unless the task is MCP/OpenCode/Code Mode work.
-- For MCP/OpenCode work, load the installed `ai_assistant` fact sheet, preserve two-tier auth and per-request ACL, and ask before config/auth/session contract changes.

package/dist/agentic/shared/ai/skills/om-create-ai-agent/references/attachments-and-overrides.md

@@ -1,12 +0,0 @@
-# Attachments, Artifacts, and Overrides
-
-Load when the agent consumes files, emits artifacts, or changes installed agents/tools.
-
-- Validate attachment type/size/count and authorize scope/record ownership before model/tool access (`artifact-authorization`).
-- Store through framework attachment/artifact services; encrypt sensitive metadata (`encrypted-storage`), use scoped/signed downloads, and define retention/cleanup (`cleanup`).
-- Keep tool results and committed traces free of raw file secrets/private transcript bodies.
-- Prefer `aiAgentExtensions` for prompt/tool/suggestion changes. Extensions apply to an existing enabled agent and cannot resurrect a disabled one.
-- Use full agent/tool override only for replacement/disable; keep map key equal to definition ID/name. Use `null` to disable.
-- Keep extension/override exports in discovered `ai-agents.ts`/`ai-tools.ts` or supported `src/modules.ts` override domains; run generation.
-
-Verify denied attachment access, expired download, cleanup, extension order, missing/disabled target, and stale structural-cache behavior.

package/dist/agentic/shared/ai/skills/om-create-ai-agent/references/module-agents-and-tools.md

@@ -1,22 +0,0 @@
-# Module Agents and Tools
-
-Load for typed in-product AI.
-
-1. Put `ai-agents.ts`/`ai-tools.ts` at module root and export typed arrays.
-2. Declare stable namespaced agent/tool IDs, module ID, label/description/prompt, allowed tools, execution mode, provider/model defaults, ACL, media, read-only/mutation posture, and loop budget.
-3. Give each tool a Zod input, scoped/wildcard ACL, correct mutation flag, bounded serializable result, and no direct cross-module table access.
-4. Write tools call `prepareMutation`; the execute callback dispatches a command with optimistic locking and side effects. No write occurs before approval.
-5. Grant features in `acl.ts`/`setup.ts`; use page-context resolvers and UI parts only when required.
-6. Run `yarn generate` and refresh structural cache when visibility/disable behavior changed.
-
-## Low-Level MCP/OpenCode Branch
-
-Use this branch only when the request explicitly targets MCP, OpenCode, or Code Mode rather than a normal typed module agent.
-
-1. Register a direct tool with `registerMcpTool(tool, { moduleId })`; use a stable MCP-compatible name, Zod `inputSchema`, non-empty `requiredFeatures` for tenant data, and a bounded serializable result.
-2. Keep the MCP server stateless per HTTP request. Production authenticates the server with `x-api-key` and the actual user/tool call with `_sessionToken`, then reloads scope and applies wildcard-aware per-tool ACL.
-3. Code Mode `search` reads generated OpenAPI; `execute` uses sandboxed `api.request()` and must preserve endpoint-level RBAC. It does not authorize undocumented or featureless mutations.
-4. Ask before OpenCode config/Docker changes, MCP auth changes, or session-token format/TTL changes. Preserve chat `sessionId` through the installed OpenCode handlers.
-5. Do not use low-level registration to bypass `defineAiTool` + `prepareMutation`; new domain mutations stay on the typed approval path.
-
-Test provider missing, denied feature/scope, invalid tool args, loop budget/stop, approval/cancel/expire/stale version, partial failure, and actual post-approval data.

package/dist/agentic/shared/ai/skills/om-create-ai-agent/references/orchestrator-agents.md

@@ -1,12 +0,0 @@
-# Orchestrator File Agents
-
-Load for file-oriented agents/subagents only after resolving the installed orchestrator guide.
-
-- Define one bounded outcome and a schema/sample that makes completion machine-checkable.
-- Declare accepted inputs/attachments, allowed tools/skills, allowed write roots, stable output/activity paths, timeout/budget, and terminal error states.
-- Put reusable procedure in an embedded thin skill; load branch references only when selected.
-- Delegate one independent task per subagent. Specify input, prohibited scope, output format, and validator; merge only validated results.
-- Keep status/resume checkpoints durable and idempotent. A natural-language claim is not completion without expected artifacts/schema.
-- Validate path containment, sanitize names, authorize artifact downloads, redact secrets, and treat repositories/prompts/files as untrusted.
-
-Test invalid outcome, missing artifact, path escape, interrupted/resumed run, subagent failure, and excessive tool/write request.

package/dist/agentic/shared/ai/skills/om-create-ai-agent/references/surface-selector.md

@@ -1,19 +0,0 @@
-# AI Surface Selector
-
-Load this reference before creating files.
-
-| Need | Surface |
-|---|---|
-| Chat or structured in-product assistant | Module `ai-agents.ts`. |
-| Reusable domain operation available to agents | Module `ai-tools.ts` / split tool packs. |
-| Explicit MCP/OpenCode registry tool | Low-level `registerMcpTool`; inspect installed `ai_assistant` facts and preserve per-request MCP/RBAC. Prefer `defineAiTool` for ordinary product-agent tools. |
-| Dynamic API exploration/execution in OpenCode | Installed Code Mode `search` + `execute`; OpenAPI discovery plus endpoint-level RBAC. |
-| Add prompt/tools/suggestions to installed agent | `aiAgentExtensions`. |
-| Replace/disable installed agent/tool | Agent/tool override or module entry override. |
-| Record-specific launcher | Widget-injected `AiChat` with scoped page context. |
-| Coding/repository automation with outcomes/files/subagents | Installed agent-orchestrator file contract. |
-| Long-lived human/retry process | Workflow engine, optionally calling an object-mode agent. |
-
-Use exact installed orchestrator context for file agents; do not conflate them with product module agents.
-
-MCP/OpenCode configuration, authentication, provider precedence, and session semantics are ask-first contract surfaces. Route them through the `ai-workflow` guide and the installed `.ai/guides/modules/ai_assistant.md` fact sheet rather than guessing.

package/dist/agentic/shared/ai/skills/om-data-model-design/references/integrity-and-concurrency.md

@@ -1,19 +0,0 @@
-# Integrity and Concurrency
-
-Load this reference for writes, locking, retries, and relation synchronization.
-
-- Dispatch domain writes through commands. Capture before/after state needed by audit and undo. In `undo`, read the stored payload with `extractUndoPayload` from `@open-mercato/shared/lib/commands/undo`, re-authorize and re-scope the reversal, make retries safe, and restore the matching cache/event/index aliases. Never ship an empty or stub `undo`; create/update/delete each reverse their persisted mutation. For custom fields you must call `buildCustomFieldResetMap` from `@open-mercato/shared/lib/commands/customFieldSnapshots`, compare the captured before/after maps, and persist its reset map on update/delete undo; `Object.assign` alone does not replace the reset-map call because cleared/removed keys also need restoration.
-- Capture custom fields with `loadCustomFieldSnapshot(em, { entityId, recordId, tenantId, organizationId })` from that same `commands/customFieldSnapshots` module. `DataEngine` has no `getCustomFields`; write them only with the single object argument `dataEngine.setCustomFields({ entityId, recordId, tenantId, organizationId, values, notify })`, never positional arguments.
-- Type each object as `CommandHandler<Input, Result>`, implement `undo: async ({ logEntry, ctx }) => { ... }`, then call `registerCommand(command)` separately; `registerCommand` returns no command object. Call `extractUndoPayload<UndoPayload>(logEntry)`. Enforce a lock with the single options object `enforceCommandOptimisticLock({ resourceKind, resourceId, current, expected, request: ctx.request })`, not positional timestamps.
-- `execute` returns `Result` directly, never `{ result, undo }`. `prepare` returns exactly `{ before: snapshot }`; `captureAfter` returns the after-snapshot value directly; `buildLog({ result, snapshots, ctx })` reads `snapshots.before`/`snapshots.after` and returns `{ resourceId: result.id, payload: { undo: ... } }`. Never invent `snapshots.snapshot` or nest payload under `{ metadata: { payload } }`, and never cast or mutate `ctx` to shuttle command-local undo state. The command bus supplies that log to `undo`. Guard a nullable `extractUndoPayload` result before reading it. `logEntry.resourceId` is nullable, so store the stable record ID in the undo payload and validate/use that required string (or explicitly guard `resourceId`) before a scoped finder call.
-- Resolve infrastructure with types: `const em = ctx.container.resolve<EntityManager>('em').fork()` and `const dataEngine = ctx.container.resolve<DataEngine>('dataEngine')`, importing `CommandRuntimeContext`, `EntityManager` from `@mikro-orm/postgresql`, and `DataEngine` from `@open-mercato/shared/lib/data/engine`. Type command helpers as `(ctx: CommandRuntimeContext)`; do not cast `ctx` or `ctx.container` to `any` and then apply generic type arguments. Trusted scope is `ctx.auth?.tenantId` plus the concrete string `ctx.selectedOrganizationId ?? ctx.organizationScope?.selectedId`; `ctx.organizationScope` is an object, never an organization ID. `ctx.auth.organizationId`, `ctx.em`, and `ctx.dataEngine` are not command runtime fields.
-- Import `withAtomicFlush` from `@open-mercato/shared/lib/commands/flush` and call `withAtomicFlush(em, [() => { em.persist(record) }, async () => { await syncRelations() }], { transaction: true })`: the second argument is an array of zero-argument phases that close over the same `EntityManager`, not one callback receiving an `em`. A phase callback must return `void`/`Promise<void>`; use a block such as `() => { em.persist(record) }`, never the expression `() => em.persist(record)` that returns the manager. Keep notifications, queues, external calls, cache invalidation, and indexing after commit, with durable compensation/recovery when those side effects can fail. Import `emitCrudSideEffects`/`emitCrudUndoSideEffects` from `@open-mercato/shared/lib/commands/helpers` and pass their documented options object (`dataEngine`, action, entity, scoped identifiers, events/indexer), not a runtime context plus invented effects descriptor.
-- MikroORM `em.create(Entity, data)` must receive all required typed fields even when entity property initializers exist: supply `created_at` and `updated_at` on create (or the entity's camel-case equivalents). Delete undo must capture a concrete full snapshot and rebuild every required field explicitly; do not pass `Partial<Entity>` to `em.create`.
-- CRUD effects use past-tense `action: 'created' | 'updated' | 'deleted'` and `identifiers: { id, tenantId, organizationId }`. Define mutable typed values (`const events: CrudEventsConfig = ...`, `const indexer: CrudIndexerConfig = ...`) as `{ module, entity, persistent?, buildPayload? }` and `{ entityType, cacheAliases? }`; do not use `as const` for mutable `cacheAliases`, or put `eventId`/loose scope fields in an effect call.
-- Give every external/queued retry a stable idempotency key and scoped uniqueness boundary.
-- Return `updatedAt` for editable records. Let `CrudForm` send the version; custom clients build the lock header and surface 409 conflicts.
-- Guard custom action/sub-resource writes at command level, normally against the aggregate parent version, with `enforceCommandOptimisticLock` from `@open-mercato/shared/lib/crud/optimistic-lock-command` (or its documented injectable guard service).
-- If a parent form mutates children, send each child's own version.
-- Persist external cursor/mapping/progress state only after the batch commits; never mark forward progress on transient failure.
-
-Tests must inject rollback between phases, race two updates, retry one operation twice, clear a nullable value, and verify side effects occur exactly after commit.

package/dist/agentic/shared/ai/skills/om-data-model-design/references/migration-workflow.md

@@ -1,13 +0,0 @@
-# Migration Workflow
-
-Load this reference whenever entity metadata changes.
-
-1. Update `data/entities.ts`, validators, commands, API projections, UI fields, encryption maps, and tests as one contract change.
-2. Run `yarn generate` when discovery/entity registration changed.
-3. Run `yarn db:generate` as a probe; inspect all SQL and snapshot changes.
-4. Remove unrelated generator churn. If scoped SQL must be written from known metadata, follow the module's existing migration style and update only its snapshot.
-5. Verify forward migration semantics, uniqueness/index names, nullable/default/backfill behavior, and safe rollback/compatibility strategy.
-6. Never modify a shipped migration. Add a new one.
-7. Ask before `yarn db:migrate`, greenfield reset, or changing a database target.
-
-Normal delivery stops after migration file/snapshot/tests; local applied state is not a PR artifact.

package/dist/agentic/shared/ai/skills/om-data-model-design/references/schema-design.md

@@ -1,14 +0,0 @@
-# Schema Design
-
-Load this reference for entity shape and relationship decisions.
-
-- Put entities in one `data/entities.ts`; use legacy decorators and explicit DB column names.
-- Prefer UUID primary keys and module-prefixed plural table/index names.
-- Tenant-owned rows require tenant/org IDs and an index beginning with the fields used by common scoped queries.
-- Editable rows require create/update timestamps and `updated_at`; append-only logs and pure junction rows may be exempt when justified.
-- Represent money as the installed monetary contract, not floating point; represent timestamps/timezones explicitly.
-- Make nullable/optional/default semantics deliberate. A clearable field accepts explicit null through validator and command.
-- Same-module relations may use ORM relations with owned/inverse sides defined. Cross-module records use scalar IDs, snapshots, events/enrichers, or `data/extensions.ts`.
-- Add scoped uniqueness and partial indexes for soft-deleted data where needed. With legacy decorators, import `Unique` from `@mikro-orm/decorators/legacy` and declare a composite constraint as `@Unique({ name, properties: [...] })`; `@Index({ unique: true })` is not a supported unique-constraint shape. Ensure retries cannot create duplicates.
-
-Use generated entity IDs rather than class-name guesses in APIs/search/widgets. App-owned modules import them from `@/.mercato/generated/entities.ids.generated`; `#generated/entities.ids.generated` is a package-internal alias and must not be copied into a standalone app. Typecheck cannot detect a valid-looking but wrong entity ID string, so also start the generated route and smoke-test the affected API against its runtime registry.

package/dist/agentic/shared/ai/skills/om-data-model-design/references/sensitive-data.md

@@ -1,10 +0,0 @@
-# Sensitive Data
-
-Load this reference when records contain PII, credentials, addresses, contact information, personal notes, or regulated data.
-
-1. Classify data sensitivity, lookup needs, retention/deletion, logs/search/export exposure, and access features.
-2. In module `encryption.ts`, import `ModuleEncryptionMap` from `@open-mercato/shared/modules/encryption`, declare `defaultEncryptionMaps: ModuleEncryptionMap[] = [{ entityId: '<module>:<entity>', fields: [{ field: '<database_field>' }] }]`, then `export default defaultEncryptionMaps` for generated registry compatibility. The entries are field-rule objects, not string names. Use a sibling `hashField` only for deterministic equality lookup. Let `TenantDataEncryptionService` populate it during encrypted writes, and query it with `lookupHashCandidates(value)` from `@open-mercato/shared/lib/encryption/aes`; use `hashForLookup(value)` from the same import only when the installed write contract explicitly requires a single keyed value. Never use raw SHA-256 for low-entropy PII such as email or phone values.
-3. Import `findWithDecryption`, `findOneWithDecryption`, or `findAndCountWithDecryption` from `@open-mercato/shared/lib/encryption/find` and make a concrete call in every implemented direct sensitive-record read path; an unused import or encryption map alone is not a decrypted read. For `makeCrudRoute`, its `entityId` + `fields` factory QueryEngine owns list decryption and `list` has no `findAndCount` override key—do not insert a decryption helper as an unsupported option. Put trusted tenant/organization constraints in each direct helper query `where` **and** pass `{ tenantId, organizationId }` as the fifth-argument decryption scope. The scope selects keys; it does not authorize or scope the ORM query. Use a null/global scope only when the installed contract explicitly permits it. Audit detail/export/search/worker/CLI and other direct ORM paths.
-4. Keep secrets out of responses, logs, errors, events, snapshots, cache keys, search documents, vector sources, and test artifacts. Exclude encrypted source values from `search.ts` field policies and indexes; index only an explicitly approved hash-only sibling for exact equality. Never sort or fuzzy-filter ciphertext.
-5. Seed/update encryption configuration through the supported command; never hand-roll KMS/AES. For every new or changed map, use an isolated initialized test tenant to reconcile the map, read back its registration, insert a fixture through the real write path, and assert the sensitive database columns are ciphertext before claiming runtime coverage.
-6. Test authorized decryption, cross-scope denial, redaction, missing keys, export/search behavior, and cleanup/retention.

package/dist/agentic/shared/ai/skills/om-eject-and-customize/references/decision-and-procedure.md

@@ -1,17 +0,0 @@
-# Eject Decision and Procedure
-
-Load before proposing ejection.
-
-Decision gate:
-
-1. State the required behavior and exact installed module/version.
-2. Reject with evidence each smaller option: response/UI extension, interceptor/guard, component/page/module override, optional package, upstream fix.
-3. List files/module size, stable IDs, migrations, direct dependencies, optional peers, and upgrade/merge ownership that the app will assume.
-4. Present rollback and ask for explicit approval.
-
-After approval:
-
-1. Run supported `yarn mercato eject <module>` from app root.
-2. Verify source landed only under `src/modules/<module>/` and registration points to `@app`.
-3. Preserve package behavior, public IDs, migrations/snapshots, scope, commands, ACL/setup, and optional-module degradation.
-4. Make the requested change, run `yarn generate`, focused/integration/build tests, and record upstream version for future merges.