ai-mesh-clean
Skill2 starsChanged 26 days agoMITView on GitHub
Use when an AI-generated, scanned or sculpted mesh must become a printable STL (watertight, scaled, above min wall, split if needed) with a licence record. Runs the mesh-capture-specialist in a fresh context through import, remesh, decimate, thickness audit, split and export, reporting before/after stats.
In this repository30 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- agent-reviewSkill
- cad-buildSkill
- couponSkill
- curriculumSkill
- dfm-reviewSkill
- fileSkill
- gddSkill
- ip-checkSkill
- led-budgetSkill
- lesson-learnedSkill
- makerworld-publishSkill
- material-pickSkill
- mcp-new-toolSkill
- ortho-sheetSkill
- pi-power-budgetSkill
- preprint-checkSkill
- promo-planSkill
- qa-gateSkill
- render-criticSkill
- research-oneSkill
- research-topicSkill
- retopoSkill
- safety-sheetSkill
- sculptSkill
- slicer-questionSkill
- style-bibleSkill
- tech-packSkill
- unit-costSkill
- verifySkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 19
Before stats with .venv\Scripts\python.exe and trimesh: faces, components, is_watertight, is_volume, volume mm^3, bbox mm, unit assumption.
- Very specific: line 21
After stats per part, same columns as step 3, plus min wall measured (state the method: Toolbox thickness check or trimesh ray-cast …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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.
Files
Overview
Clean for print: $ARGUMENTS
Output on disk, inside the project folder the maker names (see templates\project\): assets\<asset-slug>\clean\<asset>.stl (one per print part), clean.py beside them, assets\<asset-slug>\LICENCE.md, and assets\<asset-slug>\report.md with the before/after table.
When to use
Any mesh that did not come from code-CAD and is headed for one of your printers (knowledge\00-gear-inventory\): AI text- or image-to-3D output, a photogrammetry reconstruction, a Blender sculpt. Not for render topology (/retopo), not for the print orientation and support verdict (dfm-engineer after this), not for slicing (bambu-studio-expert, then /preprint-check).
Steps
- Licence gate.
LICENCE.mdmust exist with service, tier, terms URL fetched this run and dated, prompt/inputs, job id. If the mesh was generated now, fetch the service's terms or pricing page first. Not fetchable: writelicence: not verifiedin the file and add a row toresearch\inbox\ai-to-3d\<asset-slug>.md. - Min wall must arrive with its condition (material, nozzle) from
materials-scientistor a note inknowledge\03-materials\. A bare number: stop and returnNEEDS: min wall @ material, nozzle. - Before stats with
.venv\Scripts\python.exeandtrimesh: faces, components,is_watertight,is_volume, volume mm^3, bbox mm, unit assumption. Copy the raw file toraw\untouched. - Write
clean.pyfor<blender_exe> --background --python clean.py -- <raw> <target mm> <min wall mm> <out dir>(blender_exefromconfig\paths.yaml; empty means Blender is not configured, so stop and say so). One function per step: scale to target, voxel remesh (voxel size chosen against the min wall and printed), decimate to a stated face budget, recalculate normals, 3D-Print Toolbox checks (solid, intersections, thickness at the min wall; https://extensions.blender.org/add-ons/print3d-toolbox/), split into parts if thickness fails or the part exceeds the plate, export STL per part. - After stats per part, same columns as step 3, plus min wall measured (state the method: Toolbox thickness check or trimesh ray-cast sampling) and faces below it. Every
clean\*.stlmust reportis_watertight == Trueandis_volume == True. - Write
report.md: before/after table, every parameter, licence status, a split and orientation proposal fordfm-engineer. Splits follow natural part boundaries only (separate characters, accessories, bases), never a plane through a face or main body. Slicer-side notes (Fix Model, Arc Fitting on noisy meshes) are links toknowledge\02-bambu-studio\or https://wiki.bambulab.com/en/software/bambu-studio/fix-model, not restated. - Return paths, the table, licence status and numbered NEEDS MAKER questions (including service authorisation if a generation service was requested).
Common rationalizations
| Rationalization | Reality |
| "Free tier means free to sell." | Terms say what is permitted. Fetch and quote them in LICENCE.md before touching the mesh. |
| "It looks closed in the viewport." | is_watertight and is_volume from trimesh plus the Toolbox check decide, not a screenshot. |
| "1.2 mm min wall is fine for everything." | Min wall depends on material and nozzle; a bare number is rejected. |
| "I'll patch the last hole in the GUI." | clean.py is the record; a GUI fix is lost the next time the part is regenerated. Add the step to the script and re-run. |
| "Components before equals parts after." | Separate numbers, separate sentences. |
Red flags (stop and fix)
- No
LICENCE.md, or one without fetched URL and date. - A
clean\*.stlwithis_watertightfalse oris_volumefalse. - Min wall reported without method, material and nozzle.
- Output written into a folder listed in
protected_rootsinconfig\paths.yaml.
Verification
LICENCE.mdpresent with service, tier, terms URL (fetched, dated), inputs, job id.report.mdbefore/after table: faces, components, watertight, volume mm^3, bbox mm, min wall mm (method), faces below min wall, per part.- Every
clean\*.stl:is_watertight == Trueandis_volume == Trueprinted by the venv Python this session. clean.pybeside the STLs, run this session, parameters listed.- Paths returned; split/orientation proposal handed to
dfm-engineer; NEEDS MAKER questions numbered.
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
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.

