Skip to content

AutoCAD Skills v1.0.0 - Evidence-Based Mechanical CAD

Choose a tag to compare

@beiming183-cloud beiming183-cloud released this 16 Jul 09:49

Overview

AutoCAD-skills is an agent-neutral mechanical CAD skill for creating, revising, auditing, and releasing engineering drawings and models under verified Chinese GB/T conventions.

?? Codex?Claude Code ??? Agent Skills ??????????? SKILL.md,????????? references/,????????? scripts/?

Highlights

  • AutoCAD and AutoCAD MCP readiness, recovery, transactions, plotting, and DXF round-trip validation
  • Parametric 2D/3D parts, assemblies, mechanisms, gears, and complex assembly drawings
  • Geometry, topology, cross-view, interference, DRC, DFM, and manufacturing-process checks
  • GB/T drawing presentation, fits, GD&T/GPS, datum systems, tolerance stacks, and inspection planning
  • BOM/item-balloon, configuration, revision, dependency, MBD/PMI, and release-package integrity
  • Evidence requirements for kinematics, structural, thermal, fluid, fatigue, and dynamic analysis
  • Deterministic SHA-256 release manifests with explicit external-approval protection

Reliability Principles

  • A successful tool call, save, entity count, render, or exchange-file parse is not a release pass by itself.
  • Measurements/queries and visual evidence are both required for high-risk work.
  • Missing capabilities and evidence remain NOT_EVALUATED; they are never converted into optimistic assumptions.
  • Agent-generated manufacturing artifacts remain candidate after human review unless an authorized external process approves the exact revision and configuration.

Installation

Place or clone this repository as one skill directory in a runtime that supports Agent Skills, then load SKILL.md. Runtime-specific metadata is optional; the engineering workflow remains vendor-neutral.

Validation

  • Skill structure validated with the official skill validator
  • Release-manifest script tested for deterministic output, path containment, and approval gating
  • Release baseline: d59cb99007a162a7e4d33e073ae4093bdcee1515