CLAUDE.md — Supernova Vault Router

This file is a router, not a data store. It tells you which files to read for a given request. Never duplicate facts here — always link to the source.

Identity

Adam — founder of Grandview Media Group (multi-brand holding co). Email: adam@grandviewmediagroup.com.

System name: Supernova. When Adam says "Supernova," he means this Hermes-hosted Agentic OS: the vault, UI, OpenAI-backed runner, scheduled routines, and operating memory.

Reading order on every session

Read Wiki/HotCache.md for recent conversational context.
Use the routing table below to load only the files you need.
After every meaningful action, append a structured entry to Wiki/Log.md via the log-operation skill.

Routing table

When the request is about…	Read these files
Personal info about Adam	Context/MI.md
GMG business / company context	Context/Work.md
Team / personnel	Context/Team.md
Current priorities / focus	Context/Priorities.md
Active todos / pending items / "what's next?"	Wiki/Notes/_active-todos.md (maintained by EA-Orchestrator)
Brand roster	Context/Brands.md
A specific brand (overview)	`Brands/{brand-slug}/BrandContext.md`
Brand performance / audience / seasonal	`Brands/{brand-slug}/knowledge-base.md`
Brand compliance / disclaimers	`Brands/{brand-slug}/compliance.md`
Brand revenue model / metrics	`Brands/{brand-slug}/revenue-model.md`
Brand voice / pillars / CTA language	`Brands/{brand-slug}/voice-profile.md`
Direct-response copy (VSL / advertorial / PDP / hooks / email / ad copy)	Departments/Marketing/Agents/copywriter.md → RMBC II pipeline at Departments/Marketing/Skills/copywriting/ + reference at Departments/Marketing/References/rmbc/
Video storyboarding (script → scene-by-scene + `.video-gen-spec.json`)	Departments/Marketing/Agents/storyboarder.md → 5 skills at Departments/Marketing/Skills/storyboarding/ + 22 archetypes / hooks / methodology at Departments/Marketing/References/video-storyboarding/
Onboard a new brand	`/brand onboard` (slash command) → reads Skills/onboard-brand.md
Sunset a brand	`/brand offboard` (slash command) → reads Skills/offboard-brand.md
Onboard a new project	`/project onboard` (slash command) → reads Skills/onboard-project.md
Working in a department	`Departments/{dept}/_Department.md`
Running an agent	`Departments/{dept}/Agents/{agent}.md`
Running a skill	`Departments/{dept}/Skills/{skill}.md`
Cross-cutting agent (EA, Triager, etc.)	`Agents/{agent}.md`
Cross-cutting skill	`Skills/{skill}.md`
Ingesting raw input	Skills/ingest-raw.md + `Raw/` subdir
Pulling notes from Notion into the vault	`/notion ingest <url>` (slash command) → Skills/notion-ingest.md + raw provenance at `Raw/Notion/`
A specific role Adam holds at a brand	`Brands/{brand-slug}/Roles/{role-slug}.md` (ongoing scope — KPIs, team, rituals; distinct from finite projects)
Decision needed (any topic)	Departments/Decisions/_Department.md
Health / fitness / nutrition	Departments/Personal/_Department.md
Personal Fitness Command Center	Departments/Personal/Fitness/_Fitness.md → 12-week plan, WHOOP/Apple-Health-driven readiness, workouts, mobility, nutrition, supplements. UI surface: `/fitness`
WHOOP / Apple Health raw data drop	`Raw/Health/whoop/`, `Raw/Health/apple/` — see Raw/Health/whoop/README.md for schema
Personal development / learnings	Departments/Personal/Development.md → synthesized at Wiki/Learnings/
Synthesizing a podcast/book/article note	Skills/learn-from-source.md
Content library / NotebookLM-backed corpus	`/library` (slash command) → Integrations/NotebookLM.md + Wiki/Libraries/
Bulk-add YouTube content to a library	`/youtube grab <url>` (slash command) → Skills/youtube-grab.md
Static-image ad generator (Mimic)	`/mimic [analyze\|brief\|generate\|list]` → Skills/mimic-analyze-template.md · Skills/mimic-build-brief.md · Skills/mimic-generate.md. Outputs land in `Wiki/Mimic/{Briefs,Templates}/` + `supernova-ui/data/mimic/` (gallery surface at `http://localhost:3001/mimic`, read-only).
Cross-session memory / past-session recall	`/mem-search` (claude-mem plugin) → Integrations/ClaudeMem.md
Querying a library against a question	Skills/notebooklm-query.md
Decision grounded in a library	Departments/Decisions/Skills/decision-memo.md with `libraries: [...]` input
Tag conventions	_System/TagTaxonomy.md
Tool integration semantics	`Integrations/{tool}.md`
API keys / credentials / secrets	_System/SecretsRegistry.md (env var names; values in `.env`)
Model tier resolution	_System/ModelRegistry.md
YAML frontmatter conventions	_System/SchemaConventions.md
Wiki linting / health	_System/LintingRules.md
Detailed routing logic	_System/RoutingRules.md

Operating rules (hard)

Never duplicate facts here. Link to the source.
All agents and skills MUST have YAML frontmatter conforming to _System/SchemaConventions.md.
Honor model and tools declared in agent frontmatter.
Wiki nodes link via Obsidian <a class="wikilink wikilink-broken" href="#">wikilinks</a> so the graph stays connected.
Wiki/Log.md is append-only.
Files in Raw/ are inputs only — never edit them in place; synthesize into Wiki/ instead.
Brand-specific work lives under Brands/{brand}/. Cross-brand work lives under Departments/.

Model tier defaults

Resolved against _System/ModelRegistry.md:

opus → gpt-5.5 — decision engine, strategy, complex synthesis
sonnet → gpt-5.5 — default for primary agents
haiku → gpt-5.4-mini — sub-agents: research, lookup, summarize

Every agent declares model: in YAML so providers can be swapped centrally.

Top-level architecture

Holding-co (GMG)
   ├── Cross-cutting Agents/ + Skills/
   ├── Departments/ (Marketing | Operations | TaskMgmt | Personal | Decisions | Projects)
   └── Brands/ (primal-harvest, Brand-B, ...) ──── inherit shared agents; override locally

Knowledge flow:

Raw/ ──[ingest-raw]──► Wiki/  ──[lint-wiki weekly]──► healthy graph
                          │
                          └──► HotCache.md (last ~500 chars conversational ctx)
                          └──► Log.md (append-only operations)