Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Cai-aa/CAD-Agent-Hub --skill solidworks-cad-agentgit clone --depth 1 https://github.com/Cai-aa/CAD-Agent-HubWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/cai-aa/cad-agent-hub/solidworks-cad-agent)<a href="https://agentmods.dev/skills/cai-aa/cad-agent-hub/solidworks-cad-agent"><img src="https://agentmods.dev/badge/skills/cai-aa/cad-agent-hub/solidworks-cad-agent/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/cai-aa/cad-agent-hub/solidworks-cad-agent"><img src="https://agentmods.dev/badge/skills/cai-aa/cad-agent-hub/solidworks-cad-agent.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00041 | $0.01351 |
| Opus 5.5 | $0.00016 | $0.00540 |
| Sonnet 5.5 | $0.00008 | $0.00270 |
| Haiku 4.5 | $0.00004 | $0.00135 |
Grade A, and why
solidworks-cad-agent scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 90 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SolidWorks CAD Agent
This is a local Codex adaptation for Cai-aa/CAD-Agent-Hub commit
339840b5f775e81d0b18519806ccb5e59885e2a6 with the local integrity and reusable modeling-rules patches.
It is not an upstream SolidWorks release or Skill.
MCP server: solidworks-agent.
Runtime: <checkout>/MCP/Solidworks.
Python: <python-with-solidworks-agent-installed>.
Source checkout: <checkout>/MCP/Solidworks.
The installed runtime is a separate pinned copy; edits to the checkout do not update it.
Use solidworks_health_check first. The machine uses SolidWorks 2024 templates,
configured through SOLIDWORKS_PART_TEMPLATE and SOLIDWORKS_ASSEMBLY_TEMPLATE.
Use solidworks_connect(start_if_missing=false) to attach. If the requested CAD
work requires starting SolidWorks and no instance exists, use
start_if_missing=true. Keep CAD and MCP under the same Windows user and integrity level.
Before modifying an existing model, inspect its exact title/path and configuration.
Activate the intended open document, then call solidworks_bind_document with its
exact title and expected_path. Pass the returned document_token to every
existing-document write, including dimension edits, save/export, metadata and
configuration changes. New/create/open tools return their own token. Rebinding
invalidates the old token. Duplicate open titles are refused instead of guessed.
The token binds the COM document object, path, title and configuration; a detected
switch stops the write. Human UI and unrelated automation do not share the MCP lock.
Prefer solidworks_create_parametric_part for block/cylinder/tube/flange/stepped_shaft.
For custom parts read the installed schemas/feature-graph.schema.json, the
examples/cylinder-feature-graph.json, and the relevant README section. Compile
with solidworks_compile_feature_graph before execution. Native graph dimensions
are mm/degrees; the backend converts them to metres/radians. Reject non-finite
numbers before sending them. A graph creates a new part; it is not an assembly API.
Choose a fresh output path per build. Save/export refuse existing files unless
overwrite=true is explicitly supplied as part of the user's save task. Explicit
overwrites retain a backup in .swmcp-backups. Rebuild, feature errors and solid
volume are checked before saving; block/cylinder/tube builds also check analytic
volume. These checks do not establish manufacturing compliance or recover hidden
dimensions from a photo. Reopening the result adds stronger disk evidence.
Use solidworks_capture_view for visual evidence when useful.
Use one unique request_id per logical operation. Exact successful retries replay
the durable result; changed parameters or document tokens with that ID conflict.
On timeout, query solidworks_operation_result(request_id) and
solidworks_operation_status: timeout does not cancel COM. FAILED/RUNNING/UNKNOWN
mutations are never automatically replayed. After an incremental failure, inspect
the model and visible Undo group. If partial changes have been inspected and
recovered or intentionally accepted, explicitly rebind with
acknowledge_partial_changes=true and use the new token. Never blindly undo the
user's last action. Runtime reload is disabled in code; restart the idle MCP to update.
Incremental edits do not save automatically; save explicitly when the task requires
a saved deliverable. A returned ok does not establish geometric correctness.
Distinguish protocol checks, mock tests, and live SolidWorks verification in reports.
solidworks_create_two_stage_reducer builds a multi-body .sldprt, not a constrained
.sldasm. solidworks_create_output_shaft_from_dwg builds a hard-coded example,
not a general DWG reader. Concept spur gears are not involute gears; even the
involute generator needs engineering validation before manufacturing use.
For version-specific revolve and save behavior read
references/compatibility-and-lessons.md.
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 90 lines · 41 tokens per session scan A 2d766cbd2cce
solidworks-cad-agent is a skill published in the GitHub repository Cai-aa/CAD-Agent-Hub (139 stars, last pushed 4d ago), licensed MIT. It adds 41 tokens to every session and 1,351 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-11-05.
solidworks-cad-agent compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| solidworks-cad-agentCai-aa/CAD-Agent-Hub | Skill | 139 | 4d ago | MIT | A | 41 | Claude Code, Codex |
| webgl-holographic-foilnexu-io/open-design | Skill | 107k | today | Apache-2.0 | A | 41 | Claude Code, Codex |
| html-ppt-hermes-cyber-terminalnexu-io/open-design | Skill | 107k | today | Apache-2.0 | A | 53 | Claude Code, Codex |
| html-ppt-taste-brutalistnexu-io/open-design | Skill | 107k | today | Apache-2.0 | A | 78 | Claude Code, Codex |
| visual-ralphYeachan-Heo/oh-my-codex | Skill | 34k | today | MIT | A | 52 | Codex |
| glance-canonical-layoutsandroid/skills | Skill | 10k | 3d ago | Apache-2.0 | A | 56 | Claude Code |
| accessibilitymicrosoft/hve-core | Skill | 1.6k | yesterday | MIT | A | 47 | Claude Code |