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 agentmods add skills/w1ne/kernelcad-web/kernelcad-trace-from-imagenpx skills add w1ne/kernelCAD-web --skill kernelcad-trace-from-imagegit 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-trace-from-image)<a href="https://agentmods.dev/skills/w1ne/kernelcad-web/kernelcad-trace-from-image"><img src="https://agentmods.dev/badge/skills/w1ne/kernelcad-web/kernelcad-trace-from-image.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00096 | $0.02282 |
| Opus 5 | $0.00048 | $0.01141 |
| Sonnet 5 | $0.00019 | $0.00456 |
| Haiku 4.5 | $0.00010 | $0.00228 |
Grade A, and why
kernelcad-trace-from-image 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 4d 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 — 209 lines — stays where its author put it; the contents beside it link to each section on GitHub.
kernelcad-trace-from-image
Purpose
This is a coord-extractor, not a vision agent. It takes a reference photo
plus a list of requested features and returns normalized [0,1] waypoints
the agent can convert to mm via a known scale anchor and feed to a
path().spline() / path().nurbsSegment() chain.
Accuracy honesty. opencv's contour is geometrically exact on a uniform-bg
photo. Claude vision is typically 5–10% off on dense landmarks. Every
returned feature carries a per-feature confidence in [0, 1]. Treat the
output as better starting-point waypoints, not pixel-perfect coords —
the model will still need at least one render→inspect pass before finalizing.
Output replaces the burden of eyeballing dozens of (x, y) pairs off a photo;
it does not replace the spec-writer.
Inputs
.kcad.tsblockout already in place (fromblockout-model/SKILL.md).- Reference photo on disk (
./reference.jpgtypically). - A scale anchor from the Real Object Brief — at least one numeric length (mm) over a known fraction of the image width or height. Without a scale anchor the normalized output cannot be lifted to mm.
When to load this skill
Load if all three are true:
- The reference image shows an organic curve — brow, knurl, ergonomic contour, sneaker midsole, eyewear front silhouette.
- The next authoring step is a
path().spline(),path().nurbsSegment(),path().hermiteG2(), or anurbsCurve(...)ref. - Eyeballing the waypoints would take more than ~60 seconds or the
result would be unverifiable (i.e. you could not tell from looking at
your own notes whether
(12.3, 4.1)was right).
Do not load this skill for:
- Box / cylinder / rounded-rect blockouts — every primary dimension already
lives in a
param(). - References where the brief already enumerates the waypoints.
- CAD screenshots (vector geometry already, not a photo).
The trace pipeline (4 steps)
Step 1 — Decide features
Pick the smallest set of features that captures the outline you want.
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.
- 4d ago First seen · 209 lines · 96 tokens per session scan A 8d7b9654711a
kernelcad-trace-from-image is a skill published in the GitHub repository w1ne/kernelCAD-web (17 stars, last pushed 6d ago), licensed MIT. It adds 96 tokens to every session and 2,282 once invoked, about $0.0005 per session on Opus 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
rhino
Comprehensive operational skill specification for Anthropic Claude to automate, script, troubleshoot, and optimize McNeel Rhinoceros (Rhino 8), RhinoCommon API, Grasshopper visual data trees, and Rhino.Compute.
image-to-mesh-cad-workflow
End-to-end workflow for converting a 2D concept image into a parametric 3D CAD model when manual outline tracing is unreliable. Bridges the lossy "image → text-description → manual-spline-points" gap by using AI image-to-3D-mesh services (Tripo3D, Rodin, Meshy) as an intermediate step, then auditing the imported mesh…
mechanical-drafting-gbt
Create, revise, audit, validate, and present mechanical CAD using Chinese GB/T drafting conventions. Use for AutoCAD/AutoCAD MCP; parametric 2D/3D parts, complex assemblies, mechanisms, engines, gears, machined parts, consumer products, enclosures, and powered appliances; DXF/DWG/PDF/STEP and model-based definition…
modulor
2D drafting + 3D modeling via Modulor (agent-native CAD). Use when the user asks to draw floor plans, model buildings or parts, produce DXF/SVG drawings, STL/GLB meshes, IFC BIM models, or to import and modify existing DXF drawings. Requires pip install modulor.
computational-geometry
NURBS curves and surfaces, mesh geometry, boolean operations, subdivision surfaces, tessellation methods, surface analysis, and point cloud processing for AEC computational design.
ccf-visual-composer
Own rendered visuals and visual redesign from supplied content or values: plot, beautify, lay out, generate, reconstruct, and QA paper figures, visual tables, method/architecture diagrams, icons, palettes, reference-guided layout control, and editable SVG/PDF/PPTX. Use for result-table layout, color/readability…