Skills
A skill is a drop-in file that teaches an AI agent how to use Tabular Pro. Combine a skill with the Tabular Pro MCP server and any compatible agent (Claude Code, Claude Desktop, Cursor) can run an entire research project — from questionnaire design to deliverable — from a single conversation.
Why skills
The MCP server exposes the tools. A skill provides the judgment: which tools to call, in what order, with what defaults, and when to pause for confirmation. Without a skill, an agent has to guess. With one, it behaves like a senior researcher.
Available skills
Tabular Pro — Research Engine
End-to-end research workflow: questionnaire → fieldwork → cleaning → analysis → dashboards → reports. Includes sensible defaults (randomization, significance testing at 95%, rim weighting on targets) and guardrails around AI credit spend.
Installation
Claude Code
- Place the file at
.claude/skills/tabular-pro/SKILL.mdin your project (or~/.claude/skills/tabular-pro/SKILL.mdto install globally for all projects). - Make sure the
tabularproMCP server is configured — see MCP Server. - Restart Claude Code. The skill becomes available automatically; Claude invokes it when your request matches its description.
your-project/
└── .claude/
├── settings.json ← MCP server config
└── skills/
└── tabular-pro/
└── SKILL.md ← this file
Claude Desktop
- Open Settings → Developer → Skills.
- Add a new skill and paste the contents of the downloaded file.
Other agents
Any agent that supports Anthropic-style skills (Markdown with a YAML
frontmatter that has name and description) works the same way. Consult
your tool's documentation for where to place skill files.
Writing your own skill
You can extend the stock skill with organization-specific conventions — studio templates, internal naming, approved respondent panels, branded dashboard layouts. Keep the structure:
---
name: tabular-pro-acme
description: When to invoke this skill — the agent matches on this text.
---
# Body
Workflow map, defaults, guardrails, examples.
The description is the selection signal: write it so the agent picks the
right skill when the user's request matches the work it does.
Related
- MCP Server — the tools the skill orchestrates
- Claude Code Integration — detailed setup walkthrough
- API Reference — full endpoint documentation