autopilot-cad-builder
Deterministic CAD build orchestration for Autopilot Chat using openagents CAD and pane tools, including week-1 gripper flow.
Best use case
autopilot-cad-builder is best used when you need a repeatable AI agent workflow instead of a one-off prompt.
Deterministic CAD build orchestration for Autopilot Chat using openagents CAD and pane tools, including week-1 gripper flow.
Teams using autopilot-cad-builder should expect a more consistent output, faster repeated execution, less prompt rewriting.
When to use this skill
- You want a reusable workflow that can be run more than once with consistent structure.
When not to use this skill
- You only need a quick one-off answer and do not need a reusable workflow.
- You cannot install or maintain the underlying files, dependencies, or repository context.
Installation
Claude Code / Cursor / Codex
Manual Installation
- Download SKILL.md from GitHub
- Place it in
.claude/skills/autopilot-cad-builder/SKILL.mdinside your project - Restart your AI agent — it will auto-discover the skill
How autopilot-cad-builder Compares
| Feature / Agent | autopilot-cad-builder | Standard Approach |
|---|---|---|
| Platform Support | Not specified | Limited / Varies |
| Context Awareness | High | Baseline |
| Installation Complexity | Unknown | N/A |
Frequently Asked Questions
What does this skill do?
Deterministic CAD build orchestration for Autopilot Chat using openagents CAD and pane tools, including week-1 gripper flow.
Where can I find the source code?
You can find the source code on GitHub using the link provided at the top of the page.
SKILL.md Source
# Autopilot CAD Builder
Use this skill for CAD design turns from the main `Autopilot Chat` pane.
## Objective
- Keep the user in chat while building CAD in realtime.
- Use structured CAD mutations only.
- Make progress deterministic and inspectable.
- Keep snapshot truth aligned with what is visibly rendered.
## Required Tools
Use only:
- `openagents.pane.list`
- `openagents.pane.open`
- `openagents.pane.focus`
- `openagents.pane.action`
- `openagents.cad.intent`
- `openagents.cad.action`
## Operating Contract
1. Ensure CAD pane is open/focused before CAD mutations.
2. Prefer `intent_json` with typed payloads over free-form prompt edits.
3. After each mutating intent, checkpoint with snapshot/status action.
4. Keep tool sequences short and deterministic.
5. If intent parse fails, retry once with stricter `intent_json`.
6. If CAD mutation fails, return concise user-facing remediation.
7. Never claim a 2x2 grid is visible unless snapshot truth confirms it.
## Week-1 Canonical Prompt
Use this exact natural-language prompt when user intent is week-1 gripper build:
`Create a basic 2-jaw robotic gripper with a base plate, two parallel fingers, and mounting holes for a servo motor. Make it 3D-printable and parametric for easy scaling.`
## Preferred `intent_json` For Week-1
Use strict typed payloads:
```json
{
"intent": "CreateParallelJawGripperSpec",
"jaw_open_mm": 42.0,
"finger_length_mm": 65.0,
"finger_thickness_mm": 8.0,
"base_width_mm": 78.0,
"base_depth_mm": 52.0,
"base_thickness_mm": 8.0,
"servo_mount_hole_diameter_mm": 2.9,
"print_fit_mm": 0.15,
"print_clearance_mm": 0.35
}
```
For week-1 variants, require deterministic IDs in stable order:
- `variant.baseline`
- `variant.wide-jaw`
- `variant.long-reach`
- `variant.stiff-finger`
Set material per active variant with explicit `SetMaterial` calls and verify `variant_materials` map after each assignment.
## Week-1 Build Sequence
1. `openagents.pane.open` for CAD.
2. `openagents.pane.focus` for CAD.
3. `openagents.cad.intent` with strict `CreateParallelJawGripperSpec` `intent_json`.
4. `openagents.cad.intent` for `GenerateVariants` with `count=4`.
5. Cycle active variant and call `SetMaterial` for each target variant.
6. Capture snapshot truth in single layout.
7. Toggle viewport layout (`toggle_viewport_layout` or `toggle_layout`) to quad.
8. Capture snapshot truth again and confirm all variants are visible.
9. Return final summary tied to checkpoint fields, not assumptions.
## Snapshot Truth Contract
Before asserting layout/visibility, verify:
- `design_profile == "parallel_jaw_gripper"` for week-1 path.
- `viewport_layout` is `single` or `quad`.
- `visible_variant_ids` matches current rendered layout.
- `all_variants_visible` is `false` in single and `true` in quad.
- `variant_materials` map includes all four week-1 variant IDs.
## Safety Rules
- Do not invent unsupported CAD intents.
- Do not claim completion without reading a CAD snapshot/checkpoint.
- Do not claim "all four variants are showing" unless `all_variants_visible=true` and `visible_variant_ids` includes all four IDs.
- Do not use non-`openagents.*` tools for CAD pane mutation.Related Skills
autopilot-pane-control
OpenAgents desktop pane and CAD control for Codex via openagents.* tool calls.
autopilot-data-seller
Conversational seller-authoring policy for DS-first Data Market listings in OpenAgents.
autopilot-data-seller-cli
Shell-first OpenAgents DS-first Data Market packaging and publication workflow using the deterministic packaging helper, autopilotctl, and the no-window headless runtime.
autopilot-data-market-control
Typed OpenAgents DS-first Data Market tool contract for seller publication and market read-back.
neutronpay
Neutronpay MCP and SDK workflows for Lightning, stablecoin, and fiat payments.
moneydevkit
Money Dev Kit workflows for Lightning checkout and agent wallets.
mezo
Mezo integration workflows for apps, autonomous agents, and Mezo Earn operations.
maestro
Maestro Symphony blockchain query operations for OpenAgents agents, including tip freshness checks, address/UTXO/runes queries, and production-safe troubleshooting against deployed Symphony API endpoints.
l402
L402 agent commerce workflows with lnd, lnget, scoped macaroons, aperture, and MCP.
charms
Charms workflows for Bitcoin app contracts, spell proving, and UTXO asset operations.
cast
Charms CAST DEX workflows for order creation, cancellation/replacement, partial fulfillment, signing, and Bitcoin transaction verification.
blink
Bitcoin Lightning wallet for agents — balances, invoices, payments, BTC/USD swaps, QR codes, price conversion, and transaction history via the Blink API. All output is JSON.