Praxis Landing Dynamic Page Studio
Use this skill for the official public Studio in praxis-ui-landing-page. The
Studio teaches and proves governed Praxis materialization; it must not become a
second Page Builder, component registry, execution policy language, or source of
business semantics.
Canonical Boundary And Sources
The Landing project owns the public learning/workstation experience. It consumes
published @praxisui/*, runtime profiles, official recipes, Quickstart HTTP, and
canonical documentation.
Inspect:
- landing root
AGENTS.md
src/app/pages/dynamic-page-studio/**
- execution, sandbox, network-policy, document-store, analytics, i18n, and page specs
src/app/pages/dynamic-page-examples*, especially the component and specs for
the public visual-editor opener
- Core
ComponentConfigEditorDefinition and the component owner's published
configEditor metadata before assuming an eager editor implementation
src/app/data/ai-recipes/** and example metadata/catalog sync tool
e2e/dynamic-page-studio.spec.ts plus the affected recipe/remote E2E
- package versions, runtime providers, official Quickstart origin, sitemap,
LLM indexes, crawler/publication gates, and hosted smoke
Use praxis-core-component-runtime-profiles for execution-profile semantics and
Page Builder/Core composition skills for runtime documents. Landing may explain
and project those contracts but must not redefine them.
Studio Rules
- Keep authored source and runtime materialization distinct. JSON/Visual edits
produce a canonical page document; derived runtime state, capability results,
diagnostics, receipts, and selected examples do not leak back into authored
JSON unless an explicit owner operation applies them.
- Validate the complete document before preview. Unknown components, unmatched
runtime profiles, undeclared effects, unsafe URLs/renderers/hooks, invalid
composition, or missing required policy evidence fail closed with actionable
diagnostics.
- Network policy derives from matched owner profiles and the official runtime
origin. Component ids, prompts, example labels, or visible buttons never grant
network access. Block unexpected
/api/**, cross-origin, unresolved, or
undeclared operations.
- Preserve user input across JSON/Visual mode changes without maintaining two
divergent documents. Monaco paste/import is an explicit action with parse,
validation, and recovery feedback.
- Preview scope must be visible: local-only, governed remote, degraded, blocked,
or unsupported. A rendered widget is not proof that its actions or remote data
are executable.
- Curated examples are versioned evidence with family/variant identity, learning
objective, runtime requirements, guide, and validation. Do not copy stale
vendor docs or synthetic endpoints into the catalog.
- Guided journeys expose the current step, outcome, next safe action, and
progress without hiding the actual JSON/runtime surface. They may suggest
canonical operations but never keyword-route intent or fabricate success.
- Preserve accessible splitter/editor/preview behavior, keyboard access, focus,
narrow/mobile containment, localization, and reduced-motion/high-contrast
behavior.
Workflow
- Inventory the existing page document, matched profiles, network decision,
materialization state, recipe metadata, guide, and hosted dependency.
- Classify whether the need is Studio UX, missing published recipe evidence,
under-projected component metadata, or a real owner-contract gap.
- Fix the canonical owner first when profiles, component runtime, schemas,
actions, surfaces, or composition are wrong; then update the Landing
projection and official example.
- Keep package/version and Quickstart origin aligned with the documented
integration path. Do not add local aliases or ad hoc origins.
- Update sitemap/LLM/crawler/public artifacts only when the route, guide,
catalog, or published claim changes.
Public Visual Editor Resolution And Custody
Consume the existing owner configEditor definition: resolve
editor.component ?? await editor.loadComponent?.() before opening the Settings
Panel. A lazy definition is supported metadata, not a missing editor. While the
loader is pending, or if it fails, do not substitute the JSON inspector or claim
the visual editor opened; report a safe diagnostic without exposing loader details.
Capture the selected example, authored page and widget inputs from one snapshot.
Use the latest request generation plus selection and destroy guards after the
await and for the opened panel's events. Selection A→B→A must invalidate the old
request even when the final selected ID matches. Before opening, also reject a
pending resolution if the authored page changed or its draft was reset without
a selection change. The page-snapshot check belongs to pending resolution:
a legitimate Apply replaces the authored page, so a subsequent Save in that same
owned panel must still succeed. Do not reject the panel's own revision with a
permanent comparison against the pre-open page.
On replacement, selection change or destroy, close the owned panel and release
its subscriptions; on normal close, remove the owned reference/subscriptions.
Late loader results or Apply/Save events must not open or mutate a new selection.
Keep updates in the existing canonical page/envelope through the owner outputs;
do not add a parallel draft or editor contract in Landing.
Proof
Minimum local gates:
- execution/network/sandbox/document-store focused specs;
- Dynamic Page Studio component spec;
- example metadata/catalog synchronization and validation;
- landing build;
e2e/dynamic-page-studio.spec.ts for visible workflow changes;
- affected recipe/remote E2E when a public example or Quickstart integration changes.
Prove local safe JSON, governed remote execution, unmatched/unsafe profile
rejection, blocked unexpected network, JSON/Visual round-trip, invalid paste
recovery, accessible keyboard/narrow journey, and direct route reload. Hosted
claims require the hosted smoke; local mocks do not prove public integration.
For a visual-editor opener, cover eager/lazy definitions, reverse completion,
A→B→A, pending page replacement/reset, load rejection, destroy, late events and
Apply→Save in focused specs. Then use the public button in a real browser with
the target published SDK, resolve its actual editor and exercise Apply/Save in
the canonical page. Mocked metadata/Settings Panel on an installed older SDK
train does not prove the next SDK train's lazy chunk, rendering or public flow;
keep that browser/adoption gate explicitly pending until it runs.
Companion Skills
praxis-core-component-runtime-profiles: input/effect policy evidence.
praxis-core-composition-runtime and praxis-page-builder-composition: canonical page materialization.
praxis-landing-registries-sitemap-playgrounds: registry, route, sitemap, LLM and crawler publication.
praxis-landing-public-docs-contracts: public contract claims.
praxis-angular-accessibility-governance and praxis-ui-product-design: workstation accessibility and UX.
praxis-api-quickstart-cockpit-http-validation: official remote-host evidence.
1---2name: praxis-landing-dynamic-page-studio3description: "Use when implementing or auditing the public Praxis Dynamic Page Studio and its learning journey: safe operational JSON, runtime-profile execution policy, sandbox/network fail-closed behavior, JSON/Visual modes, preview/materialization state, curated recipe families, guided outcomes/progress, accessible workspace, Quickstart integration, or hosted Studio E2E. Do not use for Page Builder runtime internals alone."4metadata:5 source: https://skillmd.com/skills/codexrodrigues/praxis-landing-dynamic-page-studio6---78# Praxis Landing Dynamic Page Studio910Use this skill for the official public Studio in `praxis-ui-landing-page`. The11Studio teaches and proves governed Praxis materialization; it must not become a12second Page Builder, component registry, execution policy language, or source of13business semantics.1415## Canonical Boundary And Sources1617The Landing project owns the public learning/workstation experience. It consumes18published `@praxisui/*`, runtime profiles, official recipes, Quickstart HTTP, and19canonical documentation.2021Inspect:2223- landing root `AGENTS.md`24- `src/app/pages/dynamic-page-studio/**`25- execution, sandbox, network-policy, document-store, analytics, i18n, and page specs26- `src/app/pages/dynamic-page-examples*`, especially the component and specs for27 the public visual-editor opener28- Core `ComponentConfigEditorDefinition` and the component owner's published29 `configEditor` metadata before assuming an eager editor implementation30- `src/app/data/ai-recipes/**` and example metadata/catalog sync tool31- `e2e/dynamic-page-studio.spec.ts` plus the affected recipe/remote E2E32- package versions, runtime providers, official Quickstart origin, sitemap,33 LLM indexes, crawler/publication gates, and hosted smoke3435Use `praxis-core-component-runtime-profiles` for execution-profile semantics and36Page Builder/Core composition skills for runtime documents. Landing may explain37and project those contracts but must not redefine them.3839## Studio Rules4041- Keep authored source and runtime materialization distinct. JSON/Visual edits42 produce a canonical page document; derived runtime state, capability results,43 diagnostics, receipts, and selected examples do not leak back into authored44 JSON unless an explicit owner operation applies them.45- Validate the complete document before preview. Unknown components, unmatched46 runtime profiles, undeclared effects, unsafe URLs/renderers/hooks, invalid47 composition, or missing required policy evidence fail closed with actionable48 diagnostics.49- Network policy derives from matched owner profiles and the official runtime50 origin. Component ids, prompts, example labels, or visible buttons never grant51 network access. Block unexpected `/api/**`, cross-origin, unresolved, or52 undeclared operations.53- Preserve user input across JSON/Visual mode changes without maintaining two54 divergent documents. Monaco paste/import is an explicit action with parse,55 validation, and recovery feedback.56- Preview scope must be visible: local-only, governed remote, degraded, blocked,57 or unsupported. A rendered widget is not proof that its actions or remote data58 are executable.59- Curated examples are versioned evidence with family/variant identity, learning60 objective, runtime requirements, guide, and validation. Do not copy stale61 vendor docs or synthetic endpoints into the catalog.62- Guided journeys expose the current step, outcome, next safe action, and63 progress without hiding the actual JSON/runtime surface. They may suggest64 canonical operations but never keyword-route intent or fabricate success.65- Preserve accessible splitter/editor/preview behavior, keyboard access, focus,66 narrow/mobile containment, localization, and reduced-motion/high-contrast67 behavior.6869## Workflow70711. Inventory the existing page document, matched profiles, network decision,72 materialization state, recipe metadata, guide, and hosted dependency.732. Classify whether the need is Studio UX, missing published recipe evidence,74 under-projected component metadata, or a real owner-contract gap.753. Fix the canonical owner first when profiles, component runtime, schemas,76 actions, surfaces, or composition are wrong; then update the Landing77 projection and official example.784. Keep package/version and Quickstart origin aligned with the documented79 integration path. Do not add local aliases or ad hoc origins.805. Update sitemap/LLM/crawler/public artifacts only when the route, guide,81 catalog, or published claim changes.8283## Public Visual Editor Resolution And Custody8485Consume the existing owner `configEditor` definition: resolve86`editor.component ?? await editor.loadComponent?.()` before opening the Settings87Panel. A lazy definition is supported metadata, not a missing editor. While the88loader is pending, or if it fails, do not substitute the JSON inspector or claim89the visual editor opened; report a safe diagnostic without exposing loader details.9091Capture the selected example, authored page and widget inputs from one snapshot.92Use the latest request generation plus selection and destroy guards after the93await and for the opened panel's events. Selection A→B→A must invalidate the old94request even when the final selected ID matches. Before opening, also reject a95pending resolution if the authored page changed or its draft was reset without96a selection change. The page-snapshot check belongs to pending resolution:97a legitimate Apply replaces the authored page, so a subsequent Save in that same98owned panel must still succeed. Do not reject the panel's own revision with a99permanent comparison against the pre-open page.100101On replacement, selection change or destroy, close the owned panel and release102its subscriptions; on normal close, remove the owned reference/subscriptions.103Late loader results or Apply/Save events must not open or mutate a new selection.104Keep updates in the existing canonical page/envelope through the owner outputs;105do not add a parallel draft or editor contract in Landing.106107## Proof108109Minimum local gates:110111- execution/network/sandbox/document-store focused specs;112- Dynamic Page Studio component spec;113- example metadata/catalog synchronization and validation;114- landing build;115- `e2e/dynamic-page-studio.spec.ts` for visible workflow changes;116- affected recipe/remote E2E when a public example or Quickstart integration changes.117118Prove local safe JSON, governed remote execution, unmatched/unsafe profile119rejection, blocked unexpected network, JSON/Visual round-trip, invalid paste120recovery, accessible keyboard/narrow journey, and direct route reload. Hosted121claims require the hosted smoke; local mocks do not prove public integration.122For a visual-editor opener, cover eager/lazy definitions, reverse completion,123A→B→A, pending page replacement/reset, load rejection, destroy, late events and124Apply→Save in focused specs. Then use the public button in a real browser with125the target published SDK, resolve its actual editor and exercise Apply/Save in126the canonical page. Mocked metadata/Settings Panel on an installed older SDK127train does not prove the next SDK train's lazy chunk, rendering or public flow;128keep that browser/adoption gate explicitly pending until it runs.129130## Companion Skills131132- `praxis-core-component-runtime-profiles`: input/effect policy evidence.133- `praxis-core-composition-runtime` and `praxis-page-builder-composition`: canonical page materialization.134- `praxis-landing-registries-sitemap-playgrounds`: registry, route, sitemap, LLM and crawler publication.135- `praxis-landing-public-docs-contracts`: public contract claims.136- `praxis-angular-accessibility-governance` and `praxis-ui-product-design`: workstation accessibility and UX.137- `praxis-api-quickstart-cockpit-http-validation`: official remote-host evidence.138