Datasheet To Symbol Lib

datasheet-to-symbol-lib

A [Claude Skill](https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview) that turns a component datasheet PDF into a KiCad symbol library.

Point Claude at a datasheet and ask it to generate KiCad symbols. The skill walks through identifying every part number, extracting and verifying pinouts (including multi-unit parts and alternate/muxed pin functions) into a JPD (JSON Part Description) file, then builds the `.kicad_sym` library with the **kipart MCP server**'s `jpd2spd`, `spd2csv`, and `kipart` tools.

Why

Hand-transcribing a pin table straight into a symbol is error prone — a dropped pin, a transposed number, a missed alternate function. None of these throw an error in the `kipart` toolchain; they just produce a wrong-but-plausible symbol. This skill's core design choice is to never go straight from "read the PDF" to a symbol: everything is landed in a **JPD file** — the JSON format `kipart` itself defines — which is machine-checked and verified against the source first. JPD is the one artifact authored by AI from which `jpd2spd` generates the SPD, `spd2csv` the CSV, and `kipart` the `.kicad_sym`, so there's a single source of truth and no re-transcription step.

What's in this repo

datasheet-to-symbol-lib/
├── SKILL.md                    # The skill itself: six-step workflow, verification, failure modes
├── example.jpd                 # A sample JPD, used to smoke-test the MCP wiring
├── scripts/
│   └── verify_jpd.py           # Mechanical checker run against the JPD before building
└── references/
    └── jpd_for_datasheets.md   # Datasheet→type/style mapping and the fields this skill adds

`SKILL.md` is the entry point Claude reads when the skill triggers. `references/jpd_for_datasheets.md` is pulled in as needed rather than loaded up front, per the [progressive disclosure](https://docs.claude.com/en/docs/agents-and-tools/agent-skills/overview) pattern skills use.

The JPD schema itself is *not* kept