3d Print

3D Print Skill

Purpose

Generate physically accurate, print-ready 3D models without fabricating dimensions. Always search for real specs first. Never guess measurements.

Workflow

Step 1 — Search for real specs

First, try the local search script:

python3 scripts/search_specs.py "Canon EF 40mm f2.8 lens outer diameter"

If the script returns no concrete millimeter values (status: no_instant_answer or vague text), do NOT stop. Use your own search_web tool to query the product's official datasheet, spec page, or Wikipedia. Extract the three key numbers: outer dimension, inner dimension, height. Only proceed to Step 2 once you have real numbers with a source you can cite.

All results are cached to specs_cache.json. On a cache hit, read the cached result and extract dimensions directly — no re-search needed.

Step 2 — Decide: preset or custom script

Check if the request matches a preset:

python3 scripts/generate_model.py --preset cube     # 20mm calibration cube
python3 scripts/generate_model.py --preset tube     # hollow cylinder
python3 scripts/generate_model.py --preset bracket  # box with through-hole

If the request does NOT match a preset (e.g. "lens cap with snap-fit tabs", "hook with chamfer", "enclosure with lip", "multi-part assembly"), write a custom Python script on the fly. Use the pattern:

import sys
sys.path.insert(0, "scripts")
import manifold3d as m3d
from generate_model import export_3mf, export_stl
from pathlib import Path

# --- build geometry using manifold3d primitives ---
# See references/manifold-examples.md for API reference:
# extrusion, revolve, hull, chamfer, loft, snap-fit tabs, threads

body = m3d.Manifold.cylinder(height=8.0, radius_low=34.0, circular_segments=64)
inner = m3d.Manifold.cylinder(height=8.2, radius_low=33.0, circular_segments=64)
result = body - inner.translate([0, 0, -0.1])

# --- export ---
out = Path("output")
export_3mf(result, out / "lens_cap.3mf", name="lens_cap")
export_stl(result, out / "lens_cap.stl")

Save the script as output/<name>_build.py, run it, then proceed to Step 3. The export_3mf and export_stl functions from generate_model.py handle all file-format details — just call them with the final Manifold object.

Step 3 — Validate the mesh

python3 scripts/validate_mesh.py output/lens_cap.stl

If validation fails, do NOT just run --fix and move on. Read the specific failing checks, trace them back to the geometry construction, and fix the root cause in the build script. Common fixes:

• watertight: FAIL → a boolean operation left an open edge; add a small Z-offset (e.g. translate([0, 0, -0.1])) on the subtracted body so it protrudes past the face
• positive_volume: FAIL → normals are inverted; the outer shell may be smaller than the inner — check radii
• no_degenerate_faces: FAIL → a zero-thickness wall from a tight boolean; add at least 0.4mm clearance between faces

After fixing, re-run the build script and re-validate. Repeat until all critical checks pass.

Step 4 — Slice

Import .3mf into Bambu Studio / PrusaSlicer / Cura. Expected: 0 non-manifold errors, 0 open edges.

Spec Extraction from Cache

When specs_cache.json already contains an entry for the target part, load and extract dimensions directly instead of re-running search:

import json
cache = json.load(open("specs_cache.json"))
# find the entry whose query matches your part
# read result["answer"] or result["abstract"]
# parse out numeric values (mm) using your own reasoning
# map to: outer_diameter, inner_diameter, height, wall_thickness, etc.

Use your reasoning to map prose descriptions ("67.5mm filter thread") to the correct variable in the build script. Do not ask the user to do this mapping manually.

Reference Files

Key Constraints

• No fabricated dimensions — search first, use real numbers with a cited source
• manifold3d for all geometry — guarantees no non-manifold edges
• Custom script when no preset fits — generate_model.py is a reference, not a limit
• 3MF primary output — ISO/IEC 25422:2025, manifold-safe, Bambu/Prusa/Cura compatible
• Fix root cause on validation failure — do not rely on --fix alone

Common Errors

Install

ln -sf /path/to/3d-print-skill ~/.claude/skills/3d-print

Dependencies

pip install manifold3d trimesh numpy requests