| name | skill-router |
| description | Unified dispatcher for 88+ skills. Use when you don't know which skill to invoke — describe your task and get routed to the optimal skill, agent composition, or MCP tool.
|
| version | 1.0.0 |
| author | agentbox-claude |
| tags | ["routing","discovery","dispatcher","meta-skill"] |
| user-invocable | true |
/route — Unified Skill Dispatcher
Describe your task. Get routed to the right skill. You don't need to know 88 skills — just say what you need.
Usage
/route [describe what you need]
Examples:
/route fix the login bug and add tests/route generate a Wardley map of our infrastructure/route audit this UI for accessibility/route research competitor pricing for UK market/route harden this Linux server for SOC2/route 123 Main St, Brooklyn — site analysis and zoning/route make a podcast about our architecture
On Start
- Read the user's input — everything after
/route.
- Classify intent against the routing table below.
- If clear match → invoke that skill immediately.
- If ambiguous → ask exactly ONE clarifying question, then route.
- If multi-skill → state the composition plan, then invoke the first skill.
Routing Table
Code Development
Methodology selection (sequential phases — not competing skills):
| Phase | Route to | Key trigger |
|---|
| You have a PRD → need docs + optional build | prd2build | "I have a PRD", "generate documentation" |
| Artifact chain traceability across a sprint | bhil-methodology | "SPEC/ADR/TASK chain", "artifact traceability" |
| Systematic 5-phase development | sparc-methodology | "spec through deployment", "SPARC" |
| Multi-agent dev + quality gates (code execution) | build-with-quality | "implement", "feature", "TDD", "quality gates" |
| Truth-score only | verification-quality | "verify", "rollback safety", "confidence score" |
Debugging & diagnosis:
| If the request involves... | Route to |
|---|
| Hard bug, intermittent failure, performance regression, "can't reproduce" | build-with-quality (debugging entry — see DEBUGGING-PROTOCOL.md) |
| Stress-test a plan or design, adversarial questioning, "grill me" | build-with-quality (design interrogation — see DEBUGGING-PROTOCOL.md) |
| Requirements ambiguous before building, clarify decision tree | build-with-quality (design interrogation) |
Other code tasks:
| If the request involves... | Route to |
|---|
| Large codebase structural analysis (call graphs, diff impact) | codebase-memory (permanent project upgrade) |
| Version-specific external library docs while coding | context7 |
| Rust systems programming | rust-development |
| WASM + JS graphics interop | wasm-js |
| CUDA GPU kernel development | cuda |
| React/Next.js/TypeScript/Tailwind conventions | bencium-code-conventions |
| Validate docs against codebase | docs-alignment |
| Large codebase structure (500+ files): call graphs, architecture overview, diff blast radius | codebase-memory (permanent project upgrade) |
| Version-specific external library docs while coding, "use context7", anti-hallucination | context7 |
GitHub Operations
| If the request involves... | Route to |
|---|
| PR review with AI agents | github-code-review |
| Cut a release, changelog, deploy | github-release-management |
| Create/modify GitHub Actions workflows | github-workflow-automation |
| Issues, project boards, sprints | github-project-management |
| Cross-repo sync, org-wide automation | github-multi-repo |
Multi-Agent / Swarm
| If the request involves... | Route to |
|---|
| Parallel agents (mesh, hierarchical, adaptive) | swarm-advanced |
| Queen-led consensus, Byzantine fault tolerance | hive-mind-advanced |
| Cloud swarm on Flow Nexus | flow-nexus-swarm |
| Sequential pipeline (step N → step N+1) | stream-chain |
| Hook automation (pre/post task, session) | hooks-automation |
| Lock-free multi-agent VCS (Jujutsu) | agentic-jujutsu |
Consultants — explicit second opinion from another LLM
Use a consultant when the coordinator wants a labelled answer from a
specific external model. The response is returned with provenance so the
caller knows who said it. Specified by PRD-005 / ADR-011.
| If the request involves... | Route to |
|---|
| Code reasoning, refactor, test gen — second opinion (OpenAI Codex CLI) | codex consultant via mcp__agentbox-consultants__codex_consult |
| Long-document or codebase-wide analysis (1M-token Gemini) | gemini consultant |
| Chinese-language reasoning, low-cost second opinion (Z.AI / GLM-5.2) | zai consultant |
| Live-web research with citations (Perplexity) | perplexity consultant |
| Math + transparent chain-of-thought (DeepSeek-reasoner) | deepseek consultant |
| User says "ask ", "consult ", "second opinion from " | matching consultant |
| User says "get a second opinion" without naming | auto-consultant agent template (classifier picks) |
Manual call form (slash-command): /consult <name> "<question>" (e.g.
/consult deepseek "verify this proof of …"). Operator guide:
docs/user/consultants.md.
Research, Web, and Content
| If the request involves... | Route to |
|---|
| Live web search with citations | perplexity-research |
| Analyse/summarise specific URLs | gemini-url-context |
| YouTube transcripts, article summaries | web-summary |
| NotebookLM: notebooks, podcasts, slides, quizzes | notebooklm |
| LinkedIn profiles, jobs, companies, messaging | linkedin |
| Reddit browsing, search, user analysis | reddit |
| SEO audit, keyword research, schema markup, GEO | toprank |
| Current/version-specific docs for external library (Next.js, React, Supabase, etc.) | context7 |
Economics
| If the request involves... | Route to |
|---|
| GPU endpoint cost estimation, inference/image-gen/analytics pricing | cost-estimation |
| Agent job cost lifecycle (estimate, hold, settle, refund) | cost-estimation |
| MRC20 DREAM token buy/withdraw calculations | cost-estimation |
| Infrastructure replacement value, COCOMO-hybrid valuation | cost-estimation |
Documents and Reports
| If the request involves... | Route to |
|---|
| Research report, white paper, policy brief | report-builder |
| Academic paper, Beamer presentation, LaTeX | latex-documents |
| Diagrams (flowchart, ER, sequence, Gantt, mindmap) | mermaid-diagrams |
| Publication-quality academic figures | paperbanana |
| Isometric network/infra diagrams | fossflow |
| Strategic Wardley maps | wardley-maps |
| Validate docs against codebase | docs-alignment |
Media, 3D, and Art
| If the request involves... | Route to |
|---|
| Video/audio transcode, edit, stream | ffmpeg-processing |
| Image format conversion, resize, batch | imagemagick |
| AI image/video generation (SD, FLUX) | comfyui |
| 3D modelling and rendering | blender |
| Game development (Godot/Unity/Unreal) | game-dev |
| "Unreal Engine", "UE5", "spawn actor", "blueprint", "PIE session" | unreal-engine (60+ MCP tools for direct editor control) |
| 3D Gaussian Splatting | lichtfeld-studio |
| Blog headers, infographics, editorial art, comics | art (Nano Banana 2) |
| Real-world geography → Minecraft worlds | terracraft |
| "Make a video", full video production, explainers, trailers, TTS, avatars, podcast-to-video | open-montage |
| "Meeting recap", "live transcription", "meeting copilot", "coaching during meeting", "record this call" | echoloop |
| "Edit this video", "find the best moments", "highlight reel", "add captions", "clone voice", "lip sync", "render for TikTok" | clipcannon (51 MCP tools, local GPU) |
Browser Automation
| If the request involves... | Route to |
|---|
| Unsure which browser tool | browser-automation (meta-skill) |
| WebGPU, WebGL, GPU rendering, 3D graph testing, VisionClaw | browsercontainer (hardware GPU, native NVIDIA Vulkan, SSE at browsercontainer:8931) |
| Desktop Chrome with login state | claude --chrome (built-in) |
| Page navigation, form-fill, screenshots | browser / playwright (both via sidecar) |
| QE-grade browser: typed assertions (16 kinds), visual-diff baseline, prompt-injection scan, semantic intent finder | qe-browser (AQE fleet, Vibium engine — aqe init to install) |
| Inspect live Chromium tabs | chrome-cdp |
| Host web server from Docker | host-webserver-debug |
| Web scraping, crawling, anti-bot bypass, Cloudflare Turnstile, infra monitoring, spider framework | scrapling (MCP: 9 tools) |
AI/ML
| If the request involves... | Route to |
|---|
| PyTorch deep learning, model training | pytorch-ml |
| Cloud neural training (Flow Nexus) | flow-nexus-neural |
| Jupyter notebook experiments | jupyter-notebooks |
| Reinforcement learning agents (Decision Transformer, Q-Learning, SARSA, etc.) | agentdb-advanced (RL Plugins section) |
| Delegate reasoning to DeepSeek | deepseek-reasoning |
| Delegate coding to GPT-5.4 (simple MCP bridge) | openai-codex |
| "Consult with OpenAI", "talk to codex", "ask GPT-5.4", code review by Codex, adversarial review, rescue when stuck | codex-companion (/codex:review, /codex:rescue) |
Memory and Learning
| If the request involves... | Route to |
|---|
| Session/long-term agent memory | agentdb-memory-patterns |
| Semantic vector search, RAG, HNSW | agentdb-vector-search |
| Distributed multi-DB sync, QUIC, hybrid search | agentdb-advanced |
| Reinforcement learning plugins (9 RL algorithms) | agentdb-advanced (RL Plugins section) |
| Session context, plan tracking, blueprints | lazy-fetch |
| "What RuVector tool helps with X?" | ruvector-catalog |
Infrastructure
| If the request involves... | Route to |
|---|
| Flow Nexus platform management | flow-nexus-platform |
| Swarm performance profiling | performance-analysis |
| Creating new skills | skill-builder |
UI/UX Design
| If the request involves... | Route to |
|---|
| General UI/UX (palettes, fonts, styles) | ui-ux-pro-max-skill |
| daisyUI / Tailwind components | daisyui |
| Enterprise UX (WCAG, regulated, ask-first) | bencium-controlled-ux-designer |
| Bold creative UX, production frontend, anti-AI-slop, shadcn/Tailwind implementation | bencium-creative (--design for direction, --build for code, default=both) |
| Visual design audit / polish | design-audit |
| Typography enforcement | typography |
| AI-first relationship interfaces | relationship-design |
Software Architecture
| If the request involves... | Route to |
|---|
| First-principles design, genuinely new things | renaissance-architecture |
| Domain modelling, systems thinking | human-architect-mindset |
| Over-engineering / vanity check | vanity-engineering-review |
| Entropy vs negentropy evaluation | negentropy-lens |
| Communication style detection | adaptive-communication |
Domain-Specific
| If the request involves... | Route to |
|---|
| Geospatial GIS, maps | qgis |
| Ontology creation (Logseq → OWL2) | ontology-core |
| Ontology validation and enrichment | ontology-enrich |
| AEO (AI search citation optimisation) | bencium-aeo |
| AEC (architecture, construction, zoning, sustainability) | architecture-studio (/studio [task]) |
Security
| If the request involves... | Route to |
|---|
| Linux hardening, compliance, forensics | defense-security |
Routing Rules
Rule 1: Clear match → dispatch immediately
State which skill handles the request in one sentence, then invoke it.
Rule 2: Ambiguous → ask ONE question
If the intent could go to 2+ skills, ask exactly one clarifying question. Then route.
Rule 3: Multi-skill composition → state the plan
If the task spans multiple skills, state the sequence and invoke the first one.
Example: "Starting with perplexity-research for competitor data, then report-builder for the analysis document."
Rule 4: No match → show condensed menu
I don't have a specific skill for that. Here's what I cover:
• Code: /route [bug fix / feature / refactor]
• Research: /route [web search / URL analysis / NotebookLM]
• Economics: /route [GPU cost / agent job pricing / token conversion / valuation]
• Design: /route [UI/UX / design audit / typography]
• Docs: /route [report / LaTeX / diagrams / Wardley map]
• Media: /route [image / video / 3D / AI art]
• DevOps: /route [GitHub / CI-CD / release]
• Security: /route [hardening / compliance / audit]
• Architecture: /route [review / first-principles / entropy lens]
• AEC: /route [site planning / zoning / sustainability]
Or browse the full inventory: see SKILL-DIRECTORY.md
Rule 5: Just /route with no arguments → show condensed menu
What This Skill Does NOT Do
- It does not execute tasks. It routes to the skill that does.
- It does not override skill-internal logic.
- It does not ask more than one clarifying question.
