Personalized Storefront Render — Substrate Skill

Translate internal artifacts into client-facing surfaces in the client's domain language. Substrate, not bespoke build.

When to invoke

• User asks for "a client-facing version of X" / "translate this for Rob/Maddie/<client>" / "ELI5 / TLDR of <internal doc>".
• User wants every internal artifact (or a class of them) to project to a client storefront.
• A new client is being onboarded and needs a personalized read-surface for ongoing work.
• A specific artifact is tagged with audiences: frontmatter and the storefront has not been regenerated.

When NOT to invoke

• The artifact is governance/SOP/internal-only with no client surface (audiences: [internal] or unmtagged).
• The work is bespoke creative direction for a specific deliverable (use the relevant domain skill — Hokage bridge content, Spiral lineage substrate, etc.).
• The client surface already exists and is hand-curated outside this substrate (do not ingest after-the-fact; respect the existing artifact).

Architecture (4 layers)

LAYER 0 — PERSONA CORPUS  (~/Documents/personas/)
  {id}.md          ← prose source-of-truth (untouched by this skill)
  {id}.lexicon.yaml ← machine-readable translation table (this skill consumes)

LAYER 1 — INTERNAL ARTIFACTS  (per client repo, /docs/**)
  Existing artifacts, optionally extended with YAML frontmatter:
  audiences[], tldr, strata, client_render_mode

LAYER 2 — TRANSLATION ENGINE  (this skill)
  Composes:
    Product Domain Engine (Phase 4 logos→pathos rendering)
    8-Strata Domain Ideal-Whole (gap-map / coverage detection)
    voice-enforcement (constitution check vs persona register)
    lexicon-substitution (forbidden-terms removal, analogy mapping, ELI5/TLDR shaping)
  Output: docs/storefront/_generated/<artifact>.<persona>.client.md
   then:  docs/storefront/_curated/<artifact>.<persona>.client.md  (after human ratify)

LAYER 3 — DEPLOY SURFACE  (per-repo, per-stack adapter)
  Hokage:  /storefront/[...slug] via Next.js → rob.<domain> (link-gated)
  Spiral:  /storefront/[...slug] via Astro → maddie.<domain> (link-gated)
  Future:  static adapter ships plain HTML

Commands (CLI surface)

The skill is callable from any agent; the operator-facing form is the organvm storefront subcommand:

Invocation by agents: prefer the skill protocol (this document) over re-implementing; call the CLI for actual rendering.

Composition contract (what this skill does NOT duplicate)

This skill is a conductor. It invokes existing skills/tools rather than re-implementing them:

• Product Domain Engine — Phase 4 (rhetorical-mode rendering) is invoked for the logos→pathos transform when audience.register requires it.
• domain-ideal-whole-substrate — the 8-strata internal-magnet rules are read to determine which source globs are eligible for client projection. The gap-map stratum surfaces audit findings.
• voice-enforcement — the persona's voice_constitution field names a rule pack that voice-scorer uses to score every draft pre-ratify.
• stranger-test-protocol — the verification step "fresh agent identifies the persona's domain language without invoking PDE/ontology terms" comes from this protocol.

Do not re-implement what these skills already do. Substrate failure modes flow downstream from substrate violations.

Substrate refuses to render orphans

Every storefront-rendered artifact MUST trace bridge_to: lineage to a real internal artifact. The substrate refuses to emit _generated/*.client.md for any input where:

• No source file exists at the path implied by the artifact frontmatter, OR
• bridge_to: references a domain handle not present in the persona's lexicon, OR
• audiences[].id references a persona without a {id}.lexicon.yaml file in ~/Documents/personas/.

This rule prevents personalization theatre — making something sound like the persona without saying anything they would actually care about. Density-of-real-content is the gravity.

Lifecycle (Universal Rule #6 — everything is a loop)

internal artifact created/updated
  → audiences frontmatter present? if no → skip (default [internal])
  → operator runs `organvm storefront sync` (on demand, no daemon — Universal Rule #55)
  → render pipeline emits _generated/<artifact>.<persona>.client.md per audience entry
  → voice-scorer scores draft against persona.voice_constitution
  → IRF auto-issues STO-DRAFT-* item per unratified draft
  → operator reads + ratifies → _curated/
  → repo-native build (npm run build for Next.js, etc.) ships /storefront/* routes
  → client receives URL; reads in their language
  → client feedback enters via mailto + `storefront feedback add` → STO-FEEDBACK-* IRF items
  → next sync regenerates from updated source + lexicon

No daemons. No LaunchAgents. Pre-push git hook in opt-in repos may warn (exit 0) when client-relevant files changed without a regenerate; the warning surfaces the exact storefront sync command to run.

Phased rollout (substrate maturity)

References

• Plan: ~/.claude/plans/2026-04-25-personalized-client-storefront-substrate.md
• Schema: ~/Workspace/organvm/schema-definitions/schemas/storefront-v1.schema.json
• Canonical example: ~/Workspace/organvm/schema-definitions/examples/storefront-frontmatter-rob.yaml
• First persona lexicon: ~/Documents/personas/rob-bonavoglia.lexicon.yaml
• First repo opt-in: ~/Workspace/4444J99/hokage-chess/storefront.config.yaml
• Sibling skills: product-domain-engine (PDE Phase 4 invocation), domain-ideal-whole-substrate (8-strata gap-map), voice-enforcement (per-persona constitution checks)