research-hub-multi-ai

research-hub Multi-AI Orchestration

research-hub can be driven by any AI that can run shell commands, call MCP tools, or call REST endpoints. Use this skill to decide when the primary assistant should work directly and when to delegate long or language-specific work to Codex CLI or Gemini CLI. The underlying workspace may connect Zotero, Obsidian, and NotebookLM, or any useful two-tool subset.

Prerequisite check (do this first)

Every command this skill emits goes through the research-hub CLI (research-hub auto, research-hub plan, research-hub ask, etc.). Before producing a delegation plan, verify the CLI exists:

research-hub --version

If that command is not found, the user installed only the Claude Code marketplace plugin. Stop and tell them:

This skill orchestrates research-hub CLI commands across multiple AIs, but the CLI itself isn't installed. Please run:

pip install research-hub-pipeline
research-hub setup --persona researcher   # or analyst | humanities | internal

Then re-run your request.

Do not produce a plan that calls research-hub ... if the CLI is missing — the user can't execute it.

Stage-agnostic, character-driven routing. This skill is NOT bound to a specific research stage (discover / ingest / design / write / submit). The decision to delegate is driven by task character — token-heavy code, long-context reading, CJK prose, mechanical bulk edits — not by which stage of the research pipeline you happen to be in. A 200-page systematic-review summary in Stage 1 routes to Gemini for the same reason a zh-TW cover letter in Stage 8 does: long CJK prose that pays off Gemini's strength.

Roles

AI	Best role	Use when
Primary assistant	planning, review, domain judgment, user-facing explanation	ambiguous tasks, quality control, final synthesis
Codex CLI	Python/backend execution, tests, mechanical bulk work, crystal generation	long local processing or code-heavy workflows
Gemini CLI	long-context drafting and CJK output	Traditional Chinese summaries, bilingual briefs, long prose drafts

Decision Rules

• Start with research-hub plan "intent" for broad research requests.
• Use research-hub auto "topic" --with-crystals --llm-cli codex for long mechanical crystal generation when Codex is available.
• Use research-hub auto "topic" --with-crystals --llm-cli gemini when the user explicitly wants Chinese/Japanese/Korean output.
• Use research-hub auto "topic" --no-nlm for first-run validation before NotebookLM automation is trusted.
• Use cached crystals or research-hub ask <cluster> "question" before spending tokens on a fresh synthesis.

Commands

research-hub plan "TOPIC"
research-hub auto "TOPIC" --with-crystals
research-hub auto "TOPIC" --with-crystals --llm-cli codex
research-hub auto "TOPIC" --with-crystals --llm-cli gemini
research-hub ask <cluster> "question"

Check available CLIs:

python -c "from research_hub.auto import detect_llm_cli; print(detect_llm_cli())"

Guardrails

• Do not delegate before the workflow is clear.
• Do not fabricate citations or metadata.
• Do not assume Gemini is Chinese-only; it can be used in English, but its strongest reason to choose it over Codex is language-heavy or long-context prose.
• Do not overwrite vault notes or delete clusters without explicit user approval.