aklofas/kicad-happy: 12 hardware design skills for Claude

kicad-happy is a collection of 12 specialized skills for Claude Code and other AI agents designed to automate hardware design review tasks in KiCad projects. The repo bundles skills for schematic parsing, PCB layout analysis, SPICE simulation, EMC compliance checks, BOM management, and component sourcing from suppliers including Digi-Key, Mouser, LCSC, and JLC PCB. At 481 stars and last updated June 2026, it represents the largest open-source hardware design skill set in the agent ecosystem, but with important caveats about install friction and maintenance clarity.

What the skills actually do

Each skill is a distinct Python module that equips an agent to parse and reason about a specific domain. The “kicad” skill parses KiCad’s S-expression format (the native file format for .sch and .kicad-pcb files) into structured data without invoking KiCad itself. This allows Claude Code to traverse nets, compute voltages from feedback divider ratios, identify subcircuits, and cross-reference component values against datasheets.

The “emc” skill embeds design rules for electromagnetic compliance checking. It audits connectors for ESD protection coverage, flags unshielded signals, and suggests applicable standards (FCC Part 15, CISPR 32, CE Marking). The “spice” skill orchestrates SPICE simulation; it writes .cir netlists from the schematic and queries simulation results. Component sourcing skills (“digikey”, “mouser”, “lcsc”, “jlcpcb”, “element14”) each interface with supplier APIs or search interfaces to find parts, check stock, retrieve datasheets, and generate formatted BOM data.

The “datasheets” skill downloads PDF datasheets by part number and (optionally) extracts key parameters. The “bom” skill reconciles bill-of-materials data across schematic and PCB layers, enforces manufacturer part number (MPN) discipline, and flags missing or ambiguous component assignments. Two lighter skills, “kidoc” (KiCad documentation) and “pcbway”, handle KiCad documentation lookup and integration with PCBWay fabrication platform respectively.

Rather than listing all 12 in isolation, the README’s practice examples are more instructive. For a robot controller board with 184 components, the agent automatically traces the power tree from USB input through two voltage rails, identifies 9 motor drive MOSFETs and their drive topology, computes RC filter cutoff frequencies, validates crystal load capacitance, and reports that 19 connectors are audited but 6 are missing ESD protection. This is not generic code review; it is domain-specific reasoning grounded in component datasheets and circuit topology.

Install paths and friction

The installation experience varies significantly by target agent platform. Claude Code users have the smoothest path: marketplace plugin install with a single command. However, the README notes a known caching issue where /plugin update may not detect new versions, requiring users to manually clear cache and reinstall to get the latest code. This is a known limitation in Claude Code itself, not this repo, but it affects the practical experience of keeping kicad-happy updated.

OpenAI Codex, Google Gemini CLI, and GitHub Copilot CLI users must either use their built-in skill installers or manually clone the repo and symlink all 12 skill directories into platform-specific config paths. Windows PowerShell users need additional symbolic link commands. The install-guidance.md file (referenced but not shown in the README excerpt) purportedly covers “platform-specific quirks, known bugs, workarounds, and OS-specific issues,” suggesting that full setup may require consulting external documentation.

Running as a GitHub Action (for automated PR design reviews) or as standalone Python scripts is possible but not explicitly detailed in the README. This leaves integration into CI/CD workflows somewhat implicit.

Comparison: kicad-happy vs. alternatives

The closest comparison point is the official Anthropic skills library, though it is general-purpose and not hardware-specific. A secondary comparison is existing KiCad automation tools (command-line utilities, KiCad plugins, dedicated EDA software), though these are not agent skills.

Feature	kicad-happy	Anthropic Official Skills	Standalone KiCad CLI Tools
S-expression parsing	Yes, built-in	No	Some (e.g., KiKit)
Agent integration	Claude Code, Codex, Gemini	All Anthropic agents	Not applicable
Power tree analysis	Yes, computed from values	No	Rare / custom
EMC compliance rules	Yes, hardcoded checks	No	No
Supplier API integration	Yes, 5 suppliers	No	No
SPICE simulation	Yes, via skill	No	Yes, via KiCad native
License	MIT	Varies	Various OSS
Last update (approx)	June 2026	Current	Variable
Community size	481 stars	Official backing	Variable

The key insight: kicad-happy trades breadth (general-purpose agent tools) for depth (hardware design knowledge). If you need an agent to download a file or run a shell command, Anthropic’s skills are correct. If you need it to understand power distribution, connector protection, and BOM discipline in a specific KiCad project, kicad-happy is a prerequisite.

Scope and coverage gaps

The repo covers the major touchpoints in a hardware design workflow: schematic and PCB parsing, simulation, supplier integration, and compliance checking. Power tree analysis is granular (computing feedback divider ratios, estimating efficiency). Connector audits flag missing ESD protection explicitly. BOM lock status is tracked (percentage of components with assigned MPNs).

However, several domains are not mentioned:

• Signal integrity (timing, trace length matching, impedance control). The README mentions “signal-integrity” as a topic tag but provides no example skill or analysis showing how this is handled.
• Thermal analysis or power dissipation simulation beyond passive RC networks.
• Layout-specific checks (creepage, clearance, plane continuity, via stitching). The skill reads PCB layout data but examples focus on connector audit and BOM tracking rather than layout design rules.
• Cost estimation or supply chain risk scoring beyond BOM sourcing.
• Custom design rules or rule file format for domain-specific boards (automotive, medical, aerospace).

These are not deal-breakers for a community project, but they indicate the skills are strongest on schematic-level analysis and supplier integration, weaker on physical design and advanced simulation.

Maintenance and dependency risk

The last update date of June 2026 is current. CI badges show active testing. The repo is MIT-licensed, permitting private modification and redistribution. However, several maintenance questions arise:

1. KiCad format volatility: KiCad is under active development. The S-expression parser is tied to specific KiCad versions. The README does not specify which KiCad versions are supported, only that Python 3.10+ is required. A major KiCad update (e.g., v9.0 or v10.0) could break the parser.

2. Supplier API stability: The Digi-Key, Mouser, LCSC, and JLC PCB integration skills depend on those vendors’ public APIs or web scraping. Supplier sites change frequently. The skills could silently fail or return stale data if upstream APIs deprecate.

3. Maintainer availability: The repo is maintained by a single author (aklofas). If maintenance stops, the community would need to fork and patch it. For critical design work, this is a risk.

4. Bloat risk: 12 skills in a single repo create implicit dependencies. Installing kicad-happy means installing all 12 skills, even if you only use 3. For environments with tight package size or security review constraints, this is inefficient.

Who should use this

kicad-happy is fit for:

• Electrical engineers and hardware designers who work in KiCad and want Claude Code to spot common design errors (missing ESD protection, unvalidated passives, undersized decoupling).
• Teams building open-source hardware boards (robotics, satellites, IoT) where design reviews are frequent and cost-sensitive.
• Educators teaching hardware design who want to use AI agents as teaching assistants for design critique.

It is less suitable for:

• Engineers locked into closed EDA tools (Altium Designer, Cadence, Mentor) or who need commercial support and liability assurance.
• Projects requiring signal integrity, thermal, or electromagnetic simulation beyond passive network validation.
• Environments with strict supply chain approval and vendor-only software policies.
• Teams that cannot tolerate community-maintained dependencies or need frozen artifact versions.

Takeaways

kicad-happy fills a genuine gap: no other open-source skill repository of comparable depth exists for hardware design agents. The schematic parser and power tree analysis are credible implementations grounded in real circuit knowledge. Plugin marketplace install for Claude Code users is straightforward, though caching issues require manual workarounds.

The repo’s primary friction points are install complexity for non-Claude Code users, unclear support for different KiCad versions, and dependency on external supplier APIs that may drift. The skill set is strongest on schematic review and BOM management, weaker on advanced simulation and physical layout design. Maintenance clarity (version support, upgrade path, SLA) is absent; for mission-critical boards, this means relying on the community or maintaining a fork.

For teams already using Claude Code and KiCad, kicad-happy is worth evaluating on a single non-critical board to assess whether the agent’s feedback matches your design standards. For teams evaluating agent skills broadly, it is the clearest proof-of-concept that domain-specific agent reasoning can add real value in specialized fields like hardware design.

Further reading

• aklofas/kicad-happy GitHub repository - Source code, issues, and community discussion for the 12 skills.
• Anthropic Claude Code documentation - Official documentation on Claude Code agent capabilities and plugin marketplace.
• Awesome KiCad list - Curated index of KiCad tools and resources, where kicad-happy is mentioned.
• KiCad project documentation - Official KiCad file format and API reference for understanding S-expression schemas.
• Joanbono/awesome-kicad repository - Community-driven collection of KiCad ecosystem tools and skill repositories.