Skip to content
Back to skills

Research Cmtkdot Arc

ASecurity

Use when the user asks to research, investigate, compare approaches, or explore a topic in depth. Supports configurable depth and multi-source synthesis.

  • 666 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added September 2, 2026
research

Security analysis

A100/100

Scanned September 2, 2026

npx -y skills add majiayu000/claude-skill-registry --skill research-cmtkdot-arc --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Research Cmtkdot Arc?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Research Cmtkdot Arc
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/majiayu000-research-cmtkdot-arc-claude-skill-registry/badge)](https://www.skillsdirectory.com/skills/majiayu000-research-cmtkdot-arc-claude-skill-registry)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: research
description: Use when the user asks to research, investigate, compare approaches, or explore a topic in depth. Supports configurable depth and multi-source synthesis.
invocation: agent
---

# Research

Deep research with multi-source synthesis. Uses `multi-ai` skill for provider routing when available.

## Workflow

### Step 1: Determine Depth

Use AskUserQuestion if depth is not specified:

- **Quick**: Fast overview of key concepts (2-3 sources)
- **Standard**: Comprehensive analysis with examples (5-8 sources)
- **Deep**: Thorough research with citations and evidence (10+ sources)

If the user provides `--depth`, skip the question.

### Step 2: Provider Routing

- If `ARC_MULTI_AI=true` and providers are available, use multi-provider research via `multi-ai` skill.
- Otherwise run Claude-only research with the same structure.

### Step 3: Multi-Source Synthesis

Launch the `persona-research-synthesizer` agent with gathered context:

```
subagent_type: "arc:persona-research-synthesizer"
run_in_background: true
prompt: "Research topic: <topic>\nDepth: <depth>\nProviders: <available providers>"
```

Output a status message and end your turn. The system wakes you when the agent finishes.

### Step 4: Report Results

```
## Research Complete

**Topic**: <topic>
**Depth**: <quick|standard|deep>
**Sources Consulted**: <count>

### Key Findings
[Structured findings with evidence]

### Comparative Analysis
[Pros/cons of different approaches]

### Recommendations
[Actionable recommendations with rationale]

### Next Steps
- `/arc:core:plan` to plan implementation based on findings
- `/arc:specialized:prd` to write a PRD incorporating research
```

## Error Handling

| Scenario | Action |
|----------|--------|
| Topic too broad | Ask user to narrow scope |
| No external providers | Continue with Claude-only research |
| Low confidence findings | Flag uncertainty, recommend deeper research |

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.