agentleFS
Sign inStart free
NatchpaiNatchpai/Hardware-Copilot/.claude/skills/pcb-inspect/SKILL.md

pcb-inspect

Skill0 starsChanged 17 days agoMITView on GitHub

Inspects and explains KiCad 10 PCB layout data in read-only mode: footprints, pads, nets, tracks, vias, placement, and distances in the saved .kicad_pcb. Use when the user asks to inspect, locate, or measure a footprint, pad, net, trace, via, or board area. Targeted queries only; it extracts evidence and does not judge the layout.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 6 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read-only: do not modify PCB, project, library, or design-rule files.

    line 15
  2. Distinctive:

    Never claim a check passed without supporting data and a stated method.

    line 19
  3. Distinctive:

    Do not assume a parser, IPC connection, or local DRC capability exists.

    line 20
  4. Very specific:

    Run scripts/query_pcb.py (Python 3.10+, standard library).

    line 24
  5. Very specific:

    Not implemented (do not offer as commands): extended component, area, check (DRC is not a local check).

    line 47
  6. Distinctive:

    If structural data is insufficient, request a targeted visual inspection, not an export of the whole board.

    line 59
  7. Distinctive:

    Never put the complete board file, full track list, or unfiltered DRC report in context.

    line 78
  8. Distinctive:

    Create no reports, temporary exports, caches, or other files unless authorized.

    line 83
  9. Very specific:

    Test commands for maintainers are in references/testing.md.

    line 84

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

PCB Inspection

Scope and safety

  • Read-only: do not modify PCB, project, library, or design-rule files.
  • Operate independently of any formal review framework.
  • The saved .kicad_pcb is the source of truth; disclose that unsaved editor changes may exist.
  • Never claim a check passed without supporting data and a stated method.
  • Do not assume a parser, IPC connection, or local DRC capability exists; verify tools before using them. Never present an unimplemented mode as available.

Commands (v1.0, Stable (Scoped))

Run scripts/query_pcb.py (Python 3.10+, standard library). Resolve the script and board paths first.

python <skill dir>/scripts/query_pcb.py <command> [targets] --board <file.kicad_pcb> [--limit 20]

CommandPurpose
summaryBoard metadata, layers, footprint counts, bounded samples.
footprint <ref>Saved placement, rotation, side, library ID, UUID.
pads <ref>Pad numbers, shapes, sizes, saved net, board-space centres.
net <name>Saved footprint/pad assignments of a net.
tracks <net>Saved segments and arcs on one net, with length.
vias <net>Saved via position, size, drill, layers, type.
nearby <ref>Other footprints by distance between saved origins.
pad-distance <ref.pin> <ref.pin>Distance between two pad centres.

--limit is 1-1000. For tracks and vias, an empty quoted target ("") requests unconnected items. States such as AMBIGUOUS_REFERENCE, AMBIGUOUS, MISSING, UNCONNECTED, ASSIGNED_WITHOUT_DECLARATION, NOT_FOUND and NOT_VERIFIED are distinct results, not passes. Read references/cli.md for exact semantics when interpreting output.

Not implemented (do not offer as commands): extended component, area, check (DRC is not a local check). If no mode is given, infer the smallest useful scope.

Procedure

  1. Identify the project, target, question, and scope.
  2. Locate the target with the cheapest read-only method; prefer the CLI or native .kicad_pcb queries over loading the file.
  3. Return only relevant geometry and connectivity: footprints, pads, nets, tracks, vias, layers, applicable constraints.
  4. Expand to connected regions, stackup, or datasheet guidance only when needed; datasheet content comes from the user, official web pages or cited NotebookLM answers (no direct PDF reads).
  5. If structural data is insufficient, request a targeted visual inspection, not an export of the whole board.
  6. Report with references, net and layer names, coordinates, evidence, and unresolved checks.

Verification boundaries

  • Saved zone outlines are not filled copper; do not assume fills are current.
  • DRC results do not establish signal integrity, power integrity, EMI, thermal performance, or manufacturability. No reported violation is not proof of correctness.
  • Distance alone does not prove decoupling effectiveness. Origin and pad-centre distances are not package or copper clearances, routed length, or timing length.
  • Track lengths are single saved primitives, not routed-path or return-path length.
  • A via's geometry and shared net do not prove it contacts a given track, pad, zone, or layer. A shared net name is an assignment lookup, not copper continuity.
  • Unrecognized constructs, filled-zone state, stackup, copper contact, and anything not safely parsed remain NOT_VERIFIED.

Context and output

  • Never put the complete board file, full track list, or unfiltered DRC report in context. Limit to the target, summarize large sets, state truncation, and offer a scoped expansion.
  • Output: Target / Scope, Findings, Evidence, Limitations / NOT_VERIFIED, Recommended next checks.
  • Create no reports, temporary exports, caches, or other files unless authorized.
  • Test commands for maintainers are in references/testing.md; golden evidence is in references/2phase-synchbuck-golden-evidence.md.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free