Pcb logo

Pcb

Community
daishuge
pcb

Design a manufacturable PCB end to end — concept, schematic, layout, routing, verification, and a staged purchase — driven from Claude Code or Codex on the desktop, with EasyEDA Pro over MCP and JLC/LCSC/Taobao in a logged-in browser. Optimises for cost-effectiveness, hand-assembly, and a free prototype board.

Overview

Publisherdaishuge
Repositorypcb-skill
Skill namepcb
Stars
160
Forks
24
Bundled files
7
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 7 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by daishuge on GitHub. Read the source before you install it.

Installation

Install the Pcb AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/daishuge/pcb-skill/tree/main/skills/pcb
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/daishuge/pcb-skill.git /tmp/pcb-skill
mkdir -p .claude/skills
cp -r /tmp/pcb-skill/skills/pcb .claude/skills/pcb
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Pcb in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Pcb on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Pcb is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

PCB

Take a hardware idea to a board a human can actually order, solder and bring up.

This skill is written from a real project that went from a blank page to a released 4-layer board. Most of what is here is not PCB theory — it is the specific set of failures that theory does not catch, and the gates that catch them.

The three principles this skill optimises for

  1. 性价比 — cost-effectiveness. Every part is chosen on measured price and real availability, netted against stock the user already owns. A price nobody has read off a live page is an ESTIMATE and must be labelled as one.
  2. 好装配 — assemblable by hand. The user solders this. No QFN/DFN/exposed-pad-only packages unless the user says otherwise. Every pad reachable with an iron, in an order that does not trap a joint behind a tall part.
  3. 免费 PCB — the board should cost nothing. JLC's free-prototype coupon has hard geometric and process limits. Design inside them from the first sketch, not at the end.

Before you start — the environment must be ready

Do not begin design work until these are true. Check, do not assume.

  • EasyEDA Pro running, with the MCP bridge installed and answering (server_info). See setup/easyeda-mcp.md.
  • A browser with the agent extension, logged in to the sites the project needs: Taobao, JD, JLC/LCSC. See setup/browser-logins.md.
  • An approval watcher if the platform pops permission cards for every browser action, or the agent blocks on a card it cannot press itself. See setup/auto-approve-*.md. Carts and payment stay manual — the watcher is deliberately bypassed there.
  • A notification channel for long runs (scripts/notify/), so silence is never ambiguous.

The workflow — four phases, each with a gate

Read references/01-workflow.md for the full version. In outline:

phaseproducesgate before moving on
1 Conceptwhat it does, for whom, the hard vs soft constraintsthe user has answered every question whose two answers give different boards
2 Schematic + sourcingfull schematic, priced BOM, staged cartsnetlist assertions pass; every price VERIFIED or labelled ESTIMATE
3 Layout + routingplaced, routed, poured boardplacement is routable, not merely legal; DRC 0/0; adversarial review clear
4 Fabrication + purchaseverified Gerber, order at the pre-payment pagethe package is verified independently of the EDA that produced it

Never place an order or pay. Drive to the pre-payment page and stop.

The rules that actually caught the bugs

These are load-bearing. references/02-review.md explains each with the failure it caught.

  • A green board is not a correct board. Every phase ends with a one-pass, read-only, adversarial review by a fresh reviewer that re-derives the numbers instead of reading them. Write its termination criterion before it starts, or it will loop.
  • Measured, not inferred. "X probably does not support Y" is not a finding. If a claim decides whether work continues, prove it. Say which numbers were measured and which were computed.
  • A wait condition must answer: if this crashed right now, would my filter emit a line? Watchdogs cover four terminal states — output present, crash signature, process gone, wall-clock timeout — plus a progress-rate floor.
  • Verify the tool before trusting the result. Three separate checkers on this project silently measured the wrong thing (a cached 3D model, a bounding box instead of a pad, a merged copper label). Calibrate every geometric transform against a known asymmetric object.
  • The last honest step is the user's. Logins, CAPTCHAs, and payment are theirs. Prepare everything else so their part is two minutes.

References — load what the current phase needs

filewhen
references/01-workflow.mdalways, at the start
references/02-review.mdbefore every review, and when writing any checker
references/03-jlc-manufacturing.mdbefore fixing the board outline, stack-up or rules
references/04-sourcing.mdchoosing parts, pricing, netting against stock, staging carts
references/05-easyeda-mcp.mdbefore the first write to EasyEDA — it is a trap list
references/06-mechanical.mdbefore declaring a board assemblable; 3D, connectors, folds
references/07-bringup.mdflashing and first power-up, especially with no USB-serial adapter

Scripts

scripts/ holds the checkers this workflow depends on. They are deliberately independent of the EDA: they parse exported documents and Gerbers, so they can contradict the tool that produced them. Each directory has its own README.

  • scripts/placement/ — courtyards from real pads, via-lane widths, module-body clearance
  • scripts/routing/ — the DSN/SES autoroute chain with a four-state watchdog and a rate floor
  • scripts/verify/ — netlist assertions, Gerber parsing, drill census, 3D interference
  • scripts/notify/ — a progress relay so a long run is never silent

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Pcb AI skill do?

Design a manufacturable PCB end to end — concept, schematic, layout, routing, verification, and a staged purchase — driven from Claude Code or Codex on the desktop, with EasyEDA Pro over MCP and JLC/LCSC/Taobao in a logged-in browser. Optimises for cost-effectiveness, hand-assembly, and a free prototype board.

Why use Pcb on TypingMind?

Because you install it once and use it with any model. Pcb is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Pcb in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/daishuge/pcb-skill/tree/main/skills/pcb. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Pcb?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Pcb?

As many as you like. As long as a model supports skills, you can use Pcb with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Pcb AI skill free?

Yes. It is published on GitHub by daishuge under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇