home / skills / zpankz / mcp-skillset / research-router

research-router skill

safe

This skill routes research and investigation tasks to specialized workflows, enabling rapid deep dives, reasoning, and learning across topics.

npx playbooks add skill zpankz/mcp-skillset --skill research-router

Review the files below or copy the command above to add this skill to your agents.

Files (1)

SKILL.md

2.1 KB

---
name: research-router
description: Routes research and investigation tasks. Triggers on research, investigate, deep-dive, explore, understand, learn, study, compare, thorough, comprehensive.
---

# Research Router

Routes research, learning, and investigation tasks to specialized skills.

## Subcategories

### Deep Research
```yaml
triggers: [comprehensive, thorough, multi-source, citations, verify, deep-dive]
skills:
  - deep-research: 7-phase research protocol with Graph of Thoughts
  - mcp-skillset-workflows: Multi-skill orchestration patterns
```

### Reasoning / Thinking
```yaml
triggers: [think, reason, analyze-deeply, complex-problem, multi-step]
skills:
  - think: Cognitive enhancement toolkit
  - reason: Recursive decomposition
  - AoT: Atom of Thoughts (formal)
  - urf: Universal Reasoning Framework
```

### Learning
```yaml
triggers: [learn, understand, explain, study, compare, evaluate]
skills:
  - sc:explain: Clear explanations
  - deepstudy: Deep study workflows
  - notebooklm: Query NotebookLM
```

## Routing Decision Tree

```
research request
    │
    ├── "deep research" explicit?
    │   └── deep-research
    │
    ├── Reasoning/thinking focus?
    │   ├── Lightweight? → think
    │   ├── Formal proof? → AoT
    │   └── Multi-scale? → urf
    │
    ├── Learning focus?
    │   ├── Explanation? → sc:explain
    │   └── Study → deepstudy
    │
    └── General investigation?
        └── deep-research
```

## Progressive Loading

1. **Light Analysis**: Load `think` or `reason` first
2. **Deep Dive**: If complexity increases, add `deep-research`
3. **Formal Verification**: If proof needed, add `AoT` or `urf`

## Managed Skills

| Skill | Purpose | Trigger |
|-------|---------|---------|
| deep-research | 7-phase research | "deep research", "thorough" |
| think | Non-linear reasoning | "think", "analyze" |
| reason | Recursive decomposition | "reason", "understand" |
| AoT | Atom of Thoughts | "prove", "verify", "formal" |
| urf | Universal framework | "multi-scale", "complex" |
| sc:explain | Explanations | "explain", "describe" |

Overview

This skill routes research, investigation, and learning requests to specialized skills for concise, staged, and reliable inquiry. It selects lightweight reasoning tools for quick analysis and progressively adds deep-research or formal-verification skills when complexity or evidence needs increase. The router is designed to reduce overhead and match tasks to the most appropriate workflow.

How this skill works

The router inspects user intent for trigger words like research, investigate, learn, deep-dive, compare, and verify, then follows a decision tree to assign one or more specialized skills. It starts with light analysis skills (think, reason) and escalates to deep-research for multi-source investigations or AoT/urf for formal verification. Progressive loading minimizes resources by loading heavy skills only when required.

When to use it

You need a staged research plan that ramps from quick analysis to deep investigation.
You want multi-skill orchestration for complex, multi-step research tasks.
You require formal verification, proofs, or rigorous argument checking.
You need clear explanations or study workflows for learning and teaching.
You want evidence-backed, multi-source deep dives with citations.

Best practices

Start requests with explicit intent words like "deep research", "verify", or "explain" to improve routing accuracy.
Prefer incremental prompts: begin with a high-level question, then request deeper phases if needed.
Ask for citations or verification explicitly when you need multi-source confirmation.
Use the router for complex tasks but keep single-step fact queries to a single specialized skill for speed.
Provide constraints (scope, time horizon, depth) to guide progressive loading and avoid unnecessary escalation.

Example use cases

Exploring a contested scientific claim: route to think for initial framing, then deep-research for cross-source verification, and AoT if formal proof is needed.
Learning a new technical topic: route to sc:explain for clear summaries, then deepstudy or notebooklm for structured study material.
Comparing methodologies across papers: route to reason for decomposition, then deep-research for thorough literature aggregation and citations.
Investigating product history or regulatory changes: start with think for timeline extraction, escalate to deep-research for primary-source confirmation.
Designing a multi-step research workflow: use progressive loading to orchestrate think → deep-research → AoT as complexity increases.

FAQ

How does the router decide when to escalate to deep-research?

Escalation happens when triggers indicate multi-source verification, explicit "deep" requests, increased complexity, or when initial light analysis signals insufficient evidence.

Can I force a specific skill to run first?

Yes. Include explicit trigger words like "deep research" or name the target skill in your prompt to override default routing and load the chosen skill immediately.