Print-check a calibration cube
Published by Benchy
benchy use calibration-cube-check
Runs on Blender 5.2 with Claude Code and the Benchy MCP server. Without Blender you can still read every part of this page.
Licensed MIT. The benchy use command is documented here and not built yet; the canvas works today and will tell you what your machine is missing.
Setup
- Blender
- 5.2
- Add-ons
- None
- MCP server
- Benchy MCP
- Agent
- Claude Code
- Tokens
- Not recorded
About
The smallest complete loop Benchy can publish: a known starting file, one skill that reads measurements and proposes exactly one change, and the notes from the run. It exists to prove the shape of a template rather than to be impressive — the geometry is a 20 mm cube.
Contents
Parts by role. A part nobody has made yet stays in this list and says so.
- Native sourceThe cube as a packed .blend — not attached yet
- Viewer modelGLB derived from the source — not attached yet
- Agent skillprint-check-a-calibration-cube
- DocumentationRun notes from the measured pass
- Print fileBambu Studio project, 0.4 mm nozzle — not attached yet
- License or noticeMIT
Run notes
- Printed the cube on a 0.4 mm nozzle at 0.2 mm layers, PLA, no supports.
- First pass averaged all nine measurements before reporting, which hid a 0.09 mm Y error behind a clean X and Z. The skill now requires all nine in before any average is taken.
- Proposed one change — steps/mm on Y — and stopped. Two changes in one pass makes the next measurement unreadable.
Skill
Licensed MIT. Readable without an account, on every template.
--- name: print-check-a-calibration-cube description: Measure a printed calibration cube against its nominal size and turn the deltas into slicer changes. license: MIT --- # Print-check a calibration cube Use this when a printer has drifted and you want numbers rather than opinions. ## What you need - A 20 mm calibration cube printed on the machine under test. - Digital callipers. - Blender 5.2 with the Benchy MCP server connected, or any agent that can read and write files in the project folder. ## Steps 1. Ask the operator for three measurements per axis, taken at different heights, and record them. Do not average until all nine are in. 2. Compare each axis mean with the nominal 20.00 mm. Report the delta in mm and as a percentage, per axis, and say which axis is worst. 3. A delta under 0.05 mm on an axis is within a consumer FDM machine's normal range. Say so instead of proposing a change. 4. For a delta above 0.05 mm, propose exactly one change: steps/mm for a consistent per-axis error, or flow rate for an error that grows with wall count. Never propose both in one pass. 5. Write the measurements, the deltas and the single proposed change to run-notes.md in the project folder so the next pass can compare. ## What this does not do It does not touch printer firmware and it does not open a slicer. It produces a number and a recommendation; a human applies it.