| name | setup |
| description | Scaffold a complete knowledge system. Conversational onboarding derives configuration from your domain. Validates against 15 kernel primitives. |
| tags | ["architecture","onboarding","derivation"] |
Skill: setup
Delegates to: @ars-contexta:vault-architect
Trigger
Invoke this skill when the user wants to create a new Ars Contexta vault, initialize a knowledge system, or run first-time setup.
Behavior
Activate the vault-architect agent in setup mode. The agent runs the full derivation flow:
Screen 1 — Product Introduction
Introduce Ars Contexta: a research-backed knowledge system that derives its structure from your domain rather than imposing a generic template. Explain that the system will ask a few questions, then generate a vault tailored to how you actually think and work.
Screen 2 — Three Presets
Present the three starting points:
- Research (Zettelkasten + Cornell) — Atomic notes, heavy processing, dense schema, self/ disabled. Best for literature review, academic work, or rigorous knowledge capture.
- Personal Assistant — Mixed granularity, self/ enabled, warm personality, entity navigation. Best for life management, journaling, and personal CRM.
- Experimental — User co-designs every dimension with full transparency. Best for power users who want explicit control over all 8 configuration dimensions.
Screen 3 — What Happens Next
Explain: "I'll ask you 2-4 questions about your work, then derive 8 configuration dimensions from your answers, propose a vault structure, and generate everything you need to start."
Conversation Phase (2-4 turns)
Open with: "Tell me about what you want to track, remember, or think about."
Extract signals passively from the user's natural language. Listen for:
- Domain vocabulary (what they call things)
- Granularity preferences (atomic facts vs. rich notes)
- Collaboration signals (solo vs. team)
- Temporal patterns (daily capture vs. project-based)
- Existing tools and friction points
Derivation Phase
Map extracted signals to the 8 configuration dimensions (0.0–1.0 scale):
- Atomicity, Organization, Linking, Processing, Session, Maintenance, Search, Automation
Confidence scoring per dimension. Resolution threshold: cumulative confidence > 1.5 per dimension. For low-confidence dimensions, default to the preset baseline.
Proposal Phase
Show the user what will be generated and explain every architectural choice in terms of their stated needs. Wait for confirmation before generating.
Generation Phase
Generate the complete vault:
- Three-space folder structure (self/, {notes}/, ops/)
- Domain-native vocabulary throughout
- Templates for each note type
- Starter MOCs (hub → domain → topic hierarchy)
ops/derivation.md and ops/derivation-manifest.mdops/config.yaml with derived dimension values
Dual Context Files (always generate both)
copilot-instructions.md — GitHub Copilot CLI formatCLAUDE.md — Claude Code / OMC format
Both are derived from the same configuration but use platform-specific syntax. This ensures vault portability across AI tooling.
Validation Phase
Check all 15 kernel primitives: markdown-yaml, wiki-links, moc-hierarchy, tree-injection, description-field, topics-footer, schema-enforcement, self-space, session-rhythm, semantic-search, unique-addresses, discovery-first, operational-learning-loop, task-stack, methodology-folder.
Report PASS/FAIL per primitive. Surface any missing elements as a checklist for the user to resolve.
