Project rules and workflows for BABAT RED, a multi-space personal OS with Finance, Learning, Documents, Wardrobe, AI, Health, Dashboard, Settings, shared components, services, and utilities. Use when changing BABAT RED architecture, spaces, UI, Graphify artifacts, animation systems, or AI-agent workflows.
Resources
14Install
npx skillscat add redurbabat/babat-red Install via the SkillsCat registry.
SKILL.md
BABAT RED Project Skill
Vision
BABAT RED is a personal operating system, not a normal website. Build it as a set of focused Spaces that feel like dedicated apps while sharing one stable data layer and one coherent design language.
Primary Spaces:
- Core / Dashboard
- Finance
- Learning
- Documents
- Wardrobe
- AI
- Health
- Settings
Core Rules
- Keep the live Vanilla app stable unless the task explicitly targets it.
- Prefer scoped upgrades over big-bang migrations.
- Do not delete existing user data, localStorage schemas, IndexedDB schemas, or attachments.
- Preserve existing behavior unless the request clearly asks to change it.
- Test after code changes.
- If service worker cached live files change, update cache behavior intentionally.
- Keep untracked user files out of commits unless the user explicitly asks to include them.
Multi-Space Rules
- Treat each Space as its own app surface with its own dashboard, navigation, color accents, settings, and workflows.
- Do not show unrelated modules inside a focused Space unless it is a deliberate switcher or global system control.
- Prefer a profile or space switcher pattern for moving between Spaces.
- Keep shared primitives in shared components, services, utilities, or design tokens.
- Avoid duplicating business logic across Vanilla and React unless the migration path is documented.
Finance Space Rules
- Finance should feel like a private financial cockpit.
- Keep accounts, transactions, import, reports, investments, AI analysis, and settings grouped inside Finance.
- Do not mix Finance with school, wardrobe, documents, or electrician workflows.
- Protect user money data and avoid destructive reset actions.
- CSV import, category logic, and reporting should be service-based and testable.
Learning Space Rules
- Learning is a dedicated app for the electrician apprenticeship and the goal of moving toward grade 6.
- Include XP, level, streak, grade targets, exam mode, flashcards, AI tutor mockups, and electrician learning modules.
- Separate learning from worksite / job operations. Work time, projects, location, and material belong to Work / Baustelle, not the pure electrician learning surface.
- Electrician learning should include electrotechnics, installations, schemas, NIV / safety, tools, error analysis, ABU, and subject-specific summaries.
- Schema training should become interactive over time: component library, wire connections, validation, animated failure explanation, and repair advice.
- Do not add real AI API costs. AI tutor areas may be mockups or local-provider placeholders.
Design Style
- Use a dark, calm, premium technical style inspired by Interstellar, Christopher Nolan, Apple, Linear, Arc Browser, and control systems.
- Prefer precise panels, restrained glow, circuit lines, depth, and subtle 3D motion.
- Avoid childish colors, toy-like gradients, excessive decoration, and clutter.
- Use animation to clarify state, hierarchy, and navigation; do not animate just for noise.
- Respect reduced-motion settings.
Code Rules
- Read existing structure before editing.
- Use existing services, components, hooks, and data models when possible.
- Add abstractions only when they reduce real duplication or establish a reusable system.
- Keep React/Vite changes inside
app/unless intentionally bridging to the live root app. - Keep Vanilla root changes small and tested because it is the current live app.
- Avoid unnecessary dependencies. Prefer official, maintained packages when a dependency is justified.
- Document architecture decisions in
docs/when they affect future agents.
AI-Agent Workflow
- Start with
SKILL.md,AGENTS.md,docs/GRAPHIFY_README.md, and the relevant docs. - For structured review and planning work, use the BABAT RED GStack-lite workflows in
docs/GSTACK_WORKFLOW.mdand.claude/commands/. - Use Graphify artifacts to orient before large changes:
graphify-out/GRAPH_REPORT.mdgraphify-out/graph.jsongraphify-out/graph.html
- Inspect the real files after using the graph; the graph is an orientation aid, not source of truth.
- Make small, reviewable commits with clear summaries.
- Run the relevant tests or builds before finishing.
Never Do Automatically
- Do not wipe localStorage, IndexedDB, user files, attachments, or generated user data.
- Do not rewrite the whole app or migrate Vanilla to React without explicit approval.
- Do not deploy, rotate secrets, add paid APIs, or add cloud credentials without explicit approval.
- Do not commit unrelated user changes.
- Do not hide known test failures.
- Do not add broad tracking, analytics, or external calls without explaining the risk first.