frameshift

v0.12.0

Data-wrangling skills backed by an embedded DuckDB MCP server. Profile a dataset first, then clean it column by column.

2 skills1 MCP serverstdio

by LanternworksMIT1.5kupdated 5 months ago

Install

$ git clone https://github.com/lanternworks/frameshift

Clone into your agent client’s plugin directory — components are discovered from fixed locations per the spec.

Layout

frameshift/

├── plugin.json
├── skills/
│   ├── clean-columns/
│   │   └── SKILL.md
│   └── profile-dataset/
│       └── SKILL.md
└── mcp.json

Skills2

clean-columnsskills/clean-columns/

Normalize types, trim strings, and reconcile categorical values one column at a time, with a diff of every change.

profile-datasetskills/profile-dataset/

Summarize a dataset's shape, null rates, cardinality, and outliers before any transformation.

MCP servers1

duckdbstdio

command: frameshift-mcp
args: --readonly
cwd: ./

Manifest

plugin.json

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "frameshift",
  "version": "0.12.0",
  "description": "Data-wrangling skills backed by an embedded DuckDB MCP server. Profile a dataset first, then clean it column by column.",
  "author": {
    "name": "Lanternworks"
  },
  "repository": "https://github.com/lanternworks/frameshift",
  "license": "MIT",
  "keywords": [
    "data",
    "duckdb",
    "etl"
  ]
}