capwitf/mechanical-cad-skills/cad-evidence-ledger

Claude Code and agent skill by capwitf/mechanical-cad-skills

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Use when CAD, SolidWorks, FreeCAD, STEP, DXF, DWG, PDF, STL, 3MF, GLB, SLDPRT, SLDASM, SLDDRW, screenshots, renders, source scripts, or engineering deliverables need a verifiable evidence manifest with file paths, hashes, modified times, source provenance, checks run, and Verified/Inferred/Unverified claims.

Install

npx vskill@latest install capwitf/mechanical-cad-skills/cad-evidence-ledger

SKILL.md

CAD Evidence Ledger

Overview

A CAD result is not ready because the final answer says it is ready. Create an evidence ledger that records the actual files, source of truth, checks, and claim status behind the delivery.

Core Rule

Before final CAD reporting, produce or update an evidence ledger when files are delivered, external tooling was used, or the task has had quality complaints. The ledger must make stale files, missing formats, unchecked geometry, and overclaims visible.

Use scripts/cad_evidence_manifest.py to generate a JSON manifest for final files. The script records file existence, size, modified time, SHA256, source provenance, recorded checks, and claim categories. It does not execute external commands.

Evidence Manifest

Record these fields:

FieldMeaning
Final filesRequested outputs with path, exists, size, modified time, hash, and extension
Source of truthCAD document, script, SolidWorks/FreeCAD model, or existing drawing used to generate outputs
ToolchainFreeCAD, SolidWorks, text-to-cad, build123d, CadQuery, converter, renderer, or manual source
Checks runCommands or inspections performed; record text only, do not fake execution
Evidence filesScreenshots, renders, cropped comparisons, logs, exported previews, or open/rebuild reports
Artifact inspectioncad-artifact-inspection JSON report path and pass/partial/fail summary when produced
ClaimsVerified, Inferred, and Unverified statements
BlockersMissing tools, missing dimensions, stale files, failed opens/renders, or manual judgement needed

Script Usage

powershell
python .\.codex\skills\cad-evidence-ledger\scripts\cad_evidence_manifest.py `
--root . `
--source .\models\bracket.py `
--toolchain "FreeCAD export" `
--check "Rendered PDF page 1 nonblank" `
--verified "Final STEP exists and hash recorded" `
--unverified "Strength not calculated; loads missing" `
--output .\deliverables\cad-evidence.json `
.\deliverables\bracket.step .\deliverables\bracket.pdf

If any final file is missing and --allow-missing was not used, the script exits nonzero. Missing files in a manifest are failure evidence, not acceptable delivery.

Claim Categories

  • Verified: directly checked from the final file, rebuilt source, open/render log, measurement, hash, screenshot, or script output.
  • Inferred: plausible from source or tool behavior, but not directly checked in the final file.
  • Unverified: not checked, blocked, missing data, missing tool, or requires human engineering judgement.

Do not bury Unverified items under a success summary. If strength, fit, layout, or native SolidWorks quality was not directly checked, say so.

Required Evidence by Deliverable

DeliverableMinimum evidence
STEP/STL/3MF/GLBexists, size, mtime, hash, bounding box or open/render evidence when tools allow
DXF/DWGexists, size, mtime, hash, entity/layer/text/dimension checks when tools allow
PDF drawingexists, size, mtime, hash, page render or screenshot, title block and layout checks
SolidWorks nativeexists, size, mtime, hash, SolidWorks open/rebuild/save/export evidence when available
Source script/modelexists, hash, rerun or rebuild status, generated outputs tied to same run
Reference matchreference/output crop evidence, region checklist, comparison notes or metrics

Handoff

Use with:

  • cad-final-verification before claiming delivery is ready.
  • cad-artifact-inspection to generate read-only parser evidence for STEP/STL/DXF/DWG/PDF artifacts.
  • cad-toolchain-preflight to record local tool proof.
  • cad-engineering-sanity to label mechanical claims.
  • cad-visual-layout and cad-design-richness to record subjective gates separately from file evidence.

Common Mistakes

  • Reporting paths without checking the files exist.
  • Mixing old exports with new source.
  • Recording commands that were planned but not run.
  • Treating hash/mtime as proof of geometry correctness.
  • Marking human judgement items as Verified.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
6
7-day trend score
0

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/capwitf/mechanical-cad-skills/cad-evidence-ledger/badge)](https://verified-skill.com/skills/capwitf/mechanical-cad-skills/cad-evidence-ledger)

Something wrong with this skill or its page? Tell us.