klayoutclaw

Detect individual material layers (graphite, graphene, bottom hBN, top hBN) from their optimal source images. Use when segmenting specific materials in a van der Waals heterostructure stack from microscope images.

Agent review protocol for validating committed flake polygons in KLayout. Use after the commit step to visually inspect polygons against the microscope image, verify contour quality, and decide pass/fail/retry.

Align microscope stack images to a GDS fabrication template using lithographic marker detection. Computes image→GDS transform and commits warped image + material contours to KLayout.

Register source microscope images to the full_stack coordinate system using SIFT (same-substrate) or Chamfer+DE (cross-substrate) alignment. Use when aligning bottom_part, top_part, or other source images to the full_stack reference image for van der Waals stack detection.

Orchestrate end-to-end nanodevice design from user query through optional flake detection, material analysis, device geometry creation, routing, evaluation, and save. Device-agnostic methodology -- the agent derives physics rules from device type and available materials.

Detect and map van der Waals heterostructure flake boundaries from microscope images into KLayout polygons. Split workflow with 5 sub-skills (align, detect, combine, commit, review) for agent-orchestrated stack detection. Use this skill when the user wants to detect flakes from multiple source images, align cross-substrate images, segment hBN/graphene/graphite stacks, or run the full stack detection pipeline.

Transform detection results into the common full_stack coordinate system, build unified traces.json, and draw overlay images. Use after align and detect steps are complete.

Insert detected flake polygons into KLayout as GDS geometry using the geometry skill's add_polygon.py. Use after combine completes to commit traces.json contours as polygons on the correct layers.

Toggle KLayout layer visibility on and off for better visualization. Use this skill when the user wants to show/hide layers, toggle layer display, change layer visibility, or focus on specific layers in KLayout. Also trigger when the user says things like "hide the metal layer", "show only mesa", "turn off layer 3", or wants to isolate layers for inspection.

Run agentic E2E tests for KlayoutClaw. Gives Claude Code natural-language tasks, lets it autonomously choose MCP tools, independently verifies layout state via direct MCP calls, and judges outcomes with gpt-5-mini. Use when the user wants to run agentic tests, validate KlayoutClaw, or benchmark agent performance.

Create geometry in KLayout via MCP — rectangles, polygons, paths, cells, and cell instances. Use this skill whenever the user wants to draw, add, or create shapes/geometry/structures in KLayout, even if they don't say "geometry" explicitly. Trigger on phrases like "draw a rectangle", "add a polygon", "create a cell", "place an instance", "make a shape", or any layout/chip/mask design geometry task.

Load a reference image (microscope photo, optical image, SEM) into KLayout as a background overlay for design alignment. Use this skill whenever the user wants to load an image, add a background image, overlay a microscope photo, import a reference picture, align to a photo, or trace over an image. Also trigger when the user says "load this image", "use this photo as reference", "overlay the microscope image", or needs to position/scale/remove background images.

Import a GDS file into the current KLayout layout, flatten hierarchy, and merge into a single top cell. Use this instead of raw Layout.read() which leaves orphaned cell indices and multi-top-cell states.

Route nanodevice contacts to bonding pads with multi-window EBL support. Use this skill when the user needs to place bonding pads, route leads from device contacts to pads, set up multi-window EBL routing with different line widths per window, or add boundary connection patches between EBL write fields. Also trigger for "fan out", "route to pads", "bonding pads", "EBL windows", "multi-pass lithography", "connection patches".

Use when implementing any task using Test-Reinforced Development — separates test writing (Overseer) from implementation (Executor) with anti-cheating enforcement. Invoke with /trd <task description>.

Capture the current KLayout layout as a PNG image for visual inspection. Use this skill whenever the user wants to see, view, screenshot, preview, or visually inspect the current layout. Also trigger when the user asks "what does it look like", "show me the layout", "take a screenshot", "capture the design", or needs a visual check of their GDS geometry.