Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add w1ne/kernelCAD-web --skill kernelcad-authoringgit clone --depth 1 https://github.com/w1ne/kernelCAD-webWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/w1ne/kernelcad-web/kernelcad-authoring)<a href="https://agentmods.dev/skills/w1ne/kernelcad-web/kernelcad-authoring"><img src="https://agentmods.dev/badge/skills/w1ne/kernelcad-web/kernelcad-authoring/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/w1ne/kernelcad-web/kernelcad-authoring"><img src="https://agentmods.dev/badge/skills/w1ne/kernelcad-web/kernelcad-authoring.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Prompt Injection · line 72 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
- high Prompt Injection · line 822 Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.Fix: Audit all comments and invisible characters. Remove any instructions that direct the agent to perform unauthorized actions. Use plain, reviewable content.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00055 | $0.04140 |
| Opus 5.5 | $0.00022 | $0.01656 |
| Sonnet 5.5 | $0.00011 | $0.00828 |
| Haiku 4.5 | $0.00006 | $0.00414 |
Grade A, and why
kernelcad-authoring scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 131 lines — stays where its author put it; the contents beside it link to each section on GitHub.
kernelCAD — authoring
Write or edit .kcad.ts TypeScript models; run kernelcad evaluate <file> or kernelcad export stl|step|dxf|3mf|glb <file> -o <out>.
Single-file scripts refuse top-level import/require with feature.invalid-args; define helpers in the same file.
This file is the core: units, the rules that prevent most failures, which API serves which intent, and the finish checklist. The full reference (every signature, CLI flag, material, recipe) lives in references/*.md beside this file; over MCP pull it with lookup_authoring_skill({ section: '<id>' }) or { query: '...' } (index at the end).
Coordinate System
- Z-up, right-handed. Lengths in millimetres, angles in degrees.
box(x, y, z): corner-anchored at the origin (spans[0, x] × [0, y] × [0, z]).box(x, y, z, true)centres ALL three axes.cylinder(h, r): axis along Z, centred in X/Y, base atz = 0, top atz = h(no centred flag).sphere(r): centred on the origin.path()...close()builds a Sketch in XY;.extrude(d)goes +Z;.revolve()turns the (radial X, axial Z) profile about Z.- Return rule: the script
returns ONEShape, or one assembly viaassembly(...).model()/.solvedModel(). Only the returned value is exported, probed and measured. A script with noreturnstill evaluates on the last CREATED shape (often a cutter or helper) and gets the warningscript.no-return— treat it as a bug and add thereturn.
Conventions
The rules that prevent most failed evaluations (from the diagnostics catalogue and repair telemetry). Fix the cause, not the symptom.
- ParamRef arithmetic uses methods.
param('w', 60) + 4throwsfeature.invalid-args; writew.add(4),w.subtract(t),w.multiply(2),w.divide(4),w.negate(); also.min/.max/.abs/.clamp. Declare params at the top with units; every numeric argument must be finite (NaN/undefinedfails capture). - Holes are features, not cylinders. Use
hole/holes/cutoutinstead ofsubtract(cylinder(...)): they drill the right size (threads, counterbores, countersinks), name their faces ('wall','floor','counterbore-*','countersink-cone') and reach drawings, BOM and DFM. Holeu/vare mm offsets from the CENTRE of the entry face (top/bottom: u=+X, v=+Y;front/back: u=+X, v=+Z;left/right: u=+Y, v=+Z), not world coordinates. - Booleans need real overlap. Give unions ≥ 0.1 mm of overlap and cut tools ≥ 0.1 mm of run-out past the faces; never rely on coincident or tangent faces. A cut that misses the body is
feature.subtractive-noop— check the tool's position and axis. - One union, one connected body.
union.disconnected= an operand floats; move it until it touches.union.member-overlap= two separately finished pieces of the same stock overlap; fuse first and.finish()once, or butt the members (span - 2 * railWidth). - Separate components are named parts. Every physically distinct component is
assembly().part(name, shape); parts that touch or move get connectors and a mate ('fastened','revolute','prismatic','cylindrical','planar', ...) with limits — not rawtranslate()offsets. Loose top-level bodies emitassembly.structure.unstructured-bodies. - Sketches are one closed loop. Close the path, drop zero-length segments and duplicate points; a revolve profile stays at
x ≥ 0(feature.revolve.crosses-axis). - Fillets and chamfers fit the wall. Keep the radius below half the local wall thickness, and every target edge at least 2 × radius long (
feature.edge-feature.short-edges-skipped) — round a plate's corners in its outline (extrudeRoundedRect) rather than filleting short vertical edges (repair-oversized-fillet); fillet/chamfer/shell BEFORE the boolean that splits the face (feature.face-ref.ambiguous-after-split) and before transforms when you address faces by name. - Face names exist only where a primitive defines them. Box: all six (
top,bottom,left,right,front,back); cylinder:top/bottomonly; sphere, extruded sketches,extrudeRoundedRect, revolves and lofts: none (feature.face-ref.not-applicable). Pick those faces, and faces a cavity cut duplicated, by query:{ byNormal: 'Z' },{ byNormal: '-Z', atZ: 0 }(axis strings'X','-X','Y','-Y','Z','-Z'). - Await async helpers.
selectEdges,lib.fromSTEP,lib.fetchPart,.boundingBox(),.recenter()and.seatOnFloor()return promises (feature.async-result.missing-await). - Read the diagnostic, then repair one cause. Every diagnostic carries a
hint; follow it.recompute.input.missingmeans an upstream feature failed — callwhy_did_this_fail. Never export or hand over a script whose last evaluate failed. - Do not declare something impossible without calling
lookup_apiandlookup_cookbookfirst. kernelCAD does NURBS lofts, sweeps, boundary patches and G2 blends; it is not a polygon sculptor or a photoreal renderer. - Helical geometry: rail from
helix(...), sweep withspine: 'helix'(threads, worms) orspine: 'smooth'(wire springs). Internal threads arehole({ thread }), never a sweep..mirror(plane)= source + reflection;.reflect(plane)= reflection only.
What ships with it
7 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago Changed 741662c91ac4
- 6d ago Changed · -1,464 lines 9d223508595e
- 7d ago Changed · +20 lines bdf01b8f0924
- 8d ago Changed · +18 lines 6347f54690aa
- 9d ago Changed · +1 lines ca906fffdbeb
- 10d ago Changed · +7 lines e1895b9cb19b
- 11d ago Changed · +32 lines 37127b6e3af7
- 12d ago Changed · +3 lines ba83101e8ae4
- 13d ago Changed b981d7990640
- 14d ago First seen · 1,514 lines · 55 tokens per session scan A 9e9d2e36ce71
kernelcad-authoring is a skill published in the GitHub repository w1ne/kernelCAD-web (33 stars, last pushed today), licensed MIT. It adds 55 tokens to every session and 4,140 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
solidworks-mcp
Build, edit, inspect and export native SolidWorks models through solidworks-mcp. Use for CAD modeling, raster-to-sketch work and SolidWorks COM/MCP diagnostics. Read before calling a SolidWorks MCP tool.
add-export
Add a new subpath export to the @cyanheads/mcp-ts-core package. Use when creating a new public API surface that consumers import from a dedicated subpath (e.g., @cyanheads/mcp-ts-core/newutil).
opengis-skills
Use when AI coding assistant needs GIS/CAD/C#/AI/IoT/3D domain expertise. Index of 96 skills across 8 categories: gis (32), cad (21), csharp (9), ai (14), iot (2), 3d (6), demos (4), others (8). Tag-based search and on-demand loading for GDAL, GeoServer, MapServer, Leaflet, PROJ, PostGIS, JTS, CesiumJS, FreeCAD…
fluent-development
This skill should be used when the user asks to "build a fluent app", "create a servicenow app in typescript", or mentions "servicenow sdk", "now-sdk", "fluent", "scoped app as code", or "pro-code development" — or when the working directory contains a now.config.json or .now.ts files.
lightcad
Use when building or extending a desktop CAD/BIM application with the open-source FeiYang LightCAD platform — plugin development, custom elements, DWG/DXF I/O, 2D drafting and 3D/BIM modeling on C#/.NET. LightCAD (FeiYang Integrated Design Platform) is a C# CAD system integrating 2D design, 3D design, and BIM.
sdk
Guide users building apps, scripts, CI pipelines, or automations on top of the Cursor TypeScript SDK (@cursor/sdk). Use when the user mentions integrating, installing, or writing code against the Cursor SDK; says Agent.create, Agent.prompt, Agent.resume, agent.send, run.stream, CursorAgentError, or @cursor/sdk; asks…