agentleFS
mementomori-comementomori-co/maker-mastery/.claude/skills/preprint-check/SKILL.md

preprint-check

Skill2 starsChanged 25 days agoMITView on GitHub

Use when a 3MF is claimed ready to print and before you send it to the printer yourself. Runs print-qa in a fresh context, which reads your load record, audits the 3MF, slices with the installed Bambu Studio CLI, runs the G-code checks and returns PASS/FAIL per pre-print checklist item.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 0 are distinctive, and 3 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.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    All commands use the project Python, .venv\Scripts\python.exe (created by tools\setup.ps1), run from the repo root.

    line 16
  2. Very specific:

    Read knowledge\00-gear-inventory\current-load.md.

    line 22
  3. Very specific:

    Read templates\preprint-checklist.md and count its items (15 in the shipped template).

    line 23

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Overview

Check: $ARGUMENTS

Gate one 3MF against the physical machine state and the GUI-only slicer checks, in code. The output is the PASS/FAIL table in the reply; artifacts land in work\slice\<stem>\ (result.json, cli.log, out.3mf, plate_N.gcode). Nothing prints on a FAIL. This skill never writes to the 3MF or to knowledge\.

No AI ever starts, sends, uploads or queues a print. This skill and the print-qa agent only judge the file. You review the table and press Print in Bambu Studio yourself.

All commands use the project Python, .venv\Scripts\python.exe (created by tools\setup.ps1), run from the repo root. The Bambu Studio path comes from config\paths.yaml (bambu_studio_exe) through tools\mm_config.py.

When to use

After bambu-studio-expert hands over a 3MF, and before you print it. Also for an existing file you want judged. Not for designing or fixing the file (that is bambu-studio-expert), not for recording what is loaded (edit knowledge\00-gear-inventory\current-load.md yourself), and not for a mid-session check the lead wants on its own (that is /qa-gate, same agent).

Steps

  1. Read knowledge\00-gear-inventory\current-load.md. Missing, or no Verified: YYYY-MM-DD line: the slots-recorded item FAILs and no filament item can PASS. Continue the audit so the rest is known, but the verdict is already FAIL.
  2. Read templates\preprint-checklist.md and count its items (15 in the shipped template). The table gets exactly one row per item, in order.
  3. .venv\Scripts\python.exe tools\audit_3mf.py $ARGUMENTS. Quote the check lines you rely on.
    • Filaments in the project vs current-load.md, slot by slot (the "project filaments = physical filaments" item).
    • Grouping mode and filament map per plate (the grouping item).
    • Object bounding boxes vs the printable area of the extruder in use.
    • Support and bridge settings vs what the part spec planned.
    • More than one material family in one object, or colour painting: FAIL the one-material item unless the part spec says multi-material is the point. PLA + PETG in one object is FAIL in every case (weak bond).
    • Process preset other than the printer's default (the 0.20mm Standard profile on a 0.4 mm nozzle), an edited layer height, a Height range Modifier or variable layer height: FAIL unless the print sheet states a sourced reason; re-run with --process-reason "<reason>" so the line quotes it.
    • Stale slice (slice info, embedded G-code, md5 and project disagree): always FAIL; the fix is a re-slice with the installed CLI and a re-audit. An OK does not cover object, part, modifier or painting edits made after slicing.
  4. .venv\Scripts\python.exe tools\slice_cli.py $ARGUMENTS --debug 2 writes work\slice\<stem>\result.json, cli.log and plate_N.gcode. Read result.json (status, warning_message per plate, Support seconds per plate). Any warning is a FAIL on the "Sliced with the installed CLI" item. Support seconds must equal the plan (0 s when the spec says "no supports"). On Windows the CLI's terminal output can be lost, so an empty cli.log proves nothing; result.json and the scripts are the evidence.
  5. Per plate, each with .venv\Scripts\python.exe:
    • tools\gcode_audit.py <plate_N.gcode> (points outside the X range, support extrusion, tool changes);
    • tools\tower_bounds.py <plate_N.gcode>;
    • tools\tower_clearance.py <plate_N.gcode> --min-gap 8 (a prime-tower brim hull touching an object, or a gap under 8 mm, is FAIL; the reason is in knowledge\02-bambu-studio\grouping-and-purge.md). Any FAIL or WARN fails the X-range or prime-tower item with the measured gap or coordinate quoted, with two exceptions:
    • gcode_audit.py's "bridge spans present" WARN belongs to the display-piece overhang item. It fires on nearly every plate, because this slicer labels bridging over sparse infill "Bridge" too. Judge it with tools\overhang_gate.py (down-facing area below the threshold at that z on a visible face) and a coupon row or written prediction.
    • audit_3mf.py's "NOT VERIFIED" WARN on support Top Z belongs to the Top Z line of the display-piece item: every Top Z is "not verified" until your support-removal coupon is scored. Quote the gap that tools\support_gap_readback.py reads back instead.
  6. Every plate with support ON: .venv\Scripts\python.exe tools\interface_coverage.py <plate_N.gcode>. Exit 2 (a visible overhang face, or one side of it, under 90 % interface coverage) is a FAIL with the region's z, side and % quoted. Exit 1 (the critical-regions or small-overhang filter is on, but every candidate reads at least 90 %) is reported as "coverage not guaranteed by the filter: read-back is the proof", not a FAIL.
  7. Display pieces (anything that stands on its own): .venv\Scripts\python.exe tools\stability_check.py with the slicer grams (--gcode for parts with supports). Centre of mass outside the support polygon is FAIL. The tip-angle and contact thresholds are proposed, not verified.
  8. Purpose: read the print's PURPOSE line from the print sheet (display, stream, functional, coupon). None stated = FAIL on the visible-ratio item. For a display or stream print, run .venv\Scripts\python.exe tools\visible_ratio.py <plate_N.gcode> --out-prefix work\visible_ratio\<stem>-plate<N> on every plate with support extrusion, quote the front and iso final ratios and the 25 / 50 / 75 % ratios, and open both PNGs. Exit 2 (below 0.60, proposed default) = FAIL whatever the coverage says; exit 1 (below 0.80) = FAIL unless the sheet gives a written reason. Functional and coupon prints PASS with "not applicable: <purpose>".
  9. Tree support: for any plate with tree support, report each trunk's diameter at the plate and at z 50 mm, and the Branch diameter and Branch diameter angle from the G-code header. A long trunk wider than 12 mm at the plate with no written reason = FAIL. No tree support = "not applicable".
  10. Print sheet: it must tell you to print the gated _SLICED.3mf directly without re-slicing, say what to do if Bambu Studio forces a re-slice (compare time and grams against the gated numbers; stop on a mismatch), and carry those gated numbers from result.json. Missing instruction or numbers = FAIL on the re-slice item.
  11. Hash line: state the full sha256 of the gated _SLICED.3mf and of each plate_N.gcode, each on the same line as its path.
  12. Fill one row per checklist item, in order. The plate-face and time-and-grams items may be NEEDS MAKER when only you can confirm. One fix per FAIL. Return the table, the verdict line and the artifact paths.

Common rationalizations

RationalizationReality
"result.json says Success, skip the G-code."The CLI does not report the prime-tower clearance warning in result.json (knowledge\02-bambu-studio\cli-slicing.md). Only the scripts prove clearance.
"The maker said in chat what is loaded."Not a dated current-load.md = FAIL. A guessed filament type gets the wrong temperatures; ask or read the load file.
"Lower the gap to 7 mm, it fits."The 8 mm rule exists because the slicer's own check ignores the brim (knowledge\02-bambu-studio\grouping-and-purge.md). The bar does not move inside a check.
"Support coverage is 100 %, so it passes."Maximising one metric is never the only gate. A support cage that hides the part fails the visible-ratio item.

Red flags

  • A PASS row without a quoted script line.
  • A skipped script.
  • Any edit to the 3MF or to a file under knowledge\.
  • Any attempt to send, upload or queue the file to a printer.

Verification

  • current-load.md status stated with its Verified: date, or ABSENT.
  • audit_3mf.py, slice_cli.py, gcode_audit.py, tower_bounds.py, tower_clearance.py outputs quoted with their paths (plus interface_coverage.py, stability_check.py and visible_ratio.py where they apply).
  • Row count equals the checklist's item count; each row PASS / FAIL / NEEDS MAKER (plate face and time-and-grams only); one fix per FAIL.
  • sha256 lines present for the gated 3MF and each plate's G-code.
  • Verdict line present; artifact paths under work\slice\<stem>\ returned. Nothing prints on a FAIL.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.