// Arcanea ecosystem self-description. Use when asked about the system map, where things live, canonical locations, how to add a skill/command/agent/hook, how to absorb a repo, or what's installed. Trigger phrases: arcanea meta, system map, where does X live, add a skill, absorb a repo, what's installed, ecosystem inventory.
ACOS self-description and configuration skill. Documents how ACOS works, how to extend it, how to add new skills/commands/agents, and how to debug the hook system. Use when building new ACOS capabilities, understanding the system architecture, or onboarding to ACOS for the first time.
Generate research-grounded visuals using the InfoGenius pipeline. Use when creating infographics, diagrams, educational visuals, or any image that benefits from factual accuracy. Supports 8 visual styles (3D, technical, minimalist, photorealistic, futuristic, vintage, cartoon, standard) and 4 audience levels.
Patterns for multi-agent coordination, task decomposition, handoffs, and workflow orchestration. Best practices for building and managing agent systems.
Design and generate premium book covers using NB2 (Nano Banana 2 / Gemini 3.1 Flash Image) with deep thinking about book tension, emotion, and genre. Pulls from top book cover design principles (Chip Kidd, Peter Mendelsund, Coralie Bickford-Smith). Auto-activates when user requests book cover, cover art, cover design, or mentions generating a cover for a book project.
Claude Agentic Creator OS - Native Claude Code implementation
Build autonomous AI agents using Claude Agent SDK with computer use, tool calling, MCP integration, and production best practices for Anthropic models
| name | arcanea-meta |
| description | Arcanea ecosystem self-description. Use when asked about the system map, where things live, canonical locations, how to add a skill/command/agent/hook, how to absorb a repo, or what's installed. Trigger phrases: arcanea meta, system map, where does X live, add a skill, absorb a repo, what's installed, ecosystem inventory. |
The system describing itself with the same primitives it uses to build everything else. This skill is the load-bearing meta layer — the place to start when you don't know where something lives.
Arcanea is not one repo. It is an ecosystem of ~90 repos at github.com/frankxai, plus four installed plugins, plus a hierarchy of skills/commands/agents/hooks at three scopes:
GLOBAL (~/.claude/) ← user-scope, every project sees this
PROJECT (<repo>/.claude/) ← per-repo overrides + additions
PLUGIN (~/.claude/plugins/cache/) ← marketplace-installed (read-only)
Most failures come from forgetting which scope a thing lives in. Default rule: if it's reusable across repos, put it global. If it's Arcanea-specific, put it project. Never duplicate across both.
references/canonical-locations.md for full table)| Concern | Canonical | Why |
|---|---|---|
| Memory | ~/.claude/projects/C--Users-frank-Arcanea/memory/ | Auto-loaded via MEMORY.md every turn |
| ACOS source | ~/agentic-creator-os/ | Has FRANK_DNA.md + memory.db (irreplaceable) |
| ACOS runtime | ~/.claude/acos/ | What the harness actually invokes |
| Arcanea monorepo | ~/Arcanea/ | Web app + packages + lore + book |
| Validation contract | ~/Arcanea/CLAUDE.md § Cached-Belief Validation Protocol | Loads every turn |
| Hooks (project) | ~/Arcanea/.claude/hooks/ | 9 hooks, project-scope |
| Hooks (global) | ~/.claude/hooks/ | 3 hooks, user-scope |
| Skills | scattered (see references/installed-stack.md) | Migration ongoing |
| Source clones (for absorption) | ~/sources/ | Pre-integration staging |
When in doubt, read references/canonical-locations.md.
Update by re-running the verification commands in references/installed-stack.md. Treat the numbers as stale after 7 days.
Disk-first, memory-as-history. Any claim about current state (versions, ship status, file paths, architecture, deploy state) requires same-turn verification OR explicit "unverified, from memory:" prefix. Memory authoritative only for intent, strategy, preferences, decision history. (See references/validation-contract.md.)
Canonical wins. When two copies exist, the canonical (per canonical-locations.md) is truth. Drift copies get DEPRECATED_DELETE_AFTER_VERIFY.md markers, never silent edits.
Absorption is provenance. Claiming "absorbed pattern X from repo Y" requires entry in references/absorption-log.md with source URL, commit SHA, integrated-at path, date, verifier. No vague "we took inspiration from" allowed.
Scope before content. Before adding a skill/command/agent/hook, decide scope (global vs project vs plugin). Wrong-scope additions create the duplication that broke ACOS.
Progressive disclosure. SKILL.md stays under 3K words and contains the mental model + workflow skeleton. Depth lives in references/. Skills that violate this become unreadable monoliths (the original acos-meta did this — it teaches progressive disclosure but didn't use it).
1. Decide scope:
- Global (multi-repo use) → ~/.claude/skills/<name>/
- Project (Arcanea-only) → ~/Arcanea/.claude/skills/<name>/
2. Create SKILL.md with YAML frontmatter (name + description with trigger phrases)
3. Keep SKILL.md under 3000 words. Mental model + workflow + key principles.
4. Create references/ subdirectory for depth content.
5. Test: invoke the trigger phrase, verify activation.
1. ~/.claude/commands/<name>.md (global) or project-scope equivalent
2. YAML frontmatter: description, optional argument-hint
3. Body is a markdown workflow — no code, pure instructions
4. Reference skills/agents by name; the harness wires them
5. Test: /<name> [args]
1. ~/.claude/agents/<name>.md
2. Frontmatter: description with usage examples, model preference (optional)
3. Body = the agent's system prompt (role, capabilities, constraints, output format)
4. Invoke via Agent tool with subagent_type=<name>
1. Decide event: SessionStart | UserPromptSubmit | PreToolUse | PostToolUse | Stop | SessionEnd | Notification
2. Write ~/.claude/hooks/<name>.sh (global) or ~/Arcanea/.claude/hooks/ (project)
3. Register in settings.json (or settings.local.json) under "hooks": { "<Event>": [...] }
4. Test: trigger the event, check the hook ran (logs at ~/.arcanea/sessions/current/)
Hook count discipline: stay under 12 total. When two hooks share >30% logic, extract a shared lib.
The contract (no exceptions):
1. git clone --depth=1 <url> ~/sources/<name>/
2. Capture HEAD SHA, list 3-5 specific patterns/files you want
3. Add entry to references/absorption-log.md (source, SHA, patterns, target paths, date, verifier)
4. Integrate the patterns at the target paths (separate session/PR)
5. Update absorption-log.md status: cloned → integrated → verified
6. After integration verified, decide: keep clone for reference or rm -rf ~/sources/<name>
Vague "inspired by" or "we took ideas from" is not absorption. Either it's logged with provenance or it didn't happen.
This skill is the home for the validation contract designed in 2026-04-15 session:
.arcanea/AUTHORITY.md (deferred, Week 2)references/validation-contract.md (deferred, Week 2)CLAUDE.md § Cached-Belief Validation Protocol (Week 1 ship target)/verify-status skill → deferred Week 3Status of Week-1 ship: designed, Guardian-approved, not yet shipped to disk.
Use when:
Don't use for: domain work (code, content, design, lore). Route to domain skills/agents.
references/ecosystem-map.md — full repo inventory + spot-check freshnessreferences/installed-stack.md — verified plugin/skill/agent/command/hook counts + how to re-verifyreferences/canonical-locations.md — which copy wins for every concern, archival plan for driftreferences/absorption-log.md — strict-format provenance tablereferences/validation-contract.md — full 5-layer contract spec + rollout plan