Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Read and apply a project-specific BUILDING_SPEC.md or DIMENSIONS.md file before generating any architectural elements. Use when: generating floor plans, 3D buildings, or any spatial design where the project has defined custom dimensions, room requirements, or design standards.
.claude/skills/terminalskills-building-spec/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-15 | ✗→✓ | ▲ Improved | 59% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 397% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 70% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 51% | 0% |
Before generating any building, floor plan, or spatial design, always look for a project-specific specification file. This skill defines the workflow for finding, parsing, and applying custom dimension constraints so that every generated element matches client requirements. When no spec file exists, fall back to defaults from the architectural-dimensions skill.
Search the project root (and parent directories) for any of these filenames:
BUILDING_SPEC.mdDIMENSIONS.mdbuilding-spec.mddimensions.mdIf found, load it and treat it as ground truth for all dimensions.
Extract structured parameters from the spec file:
When generating building geometry:
After generation, check every element:
Projects should create this file in the root directory:
markdown# Building Dimensions Specification ## Project Information - **Project name:** [Your Project Name] - **Standard:** Metric (meters) | Imperial (feet/inches) - **Building type:** Residential | Commercial | Mixed-use ## Floor Heights - Ground floor: 2.7m - Upper floors: 2.7m - Basement: 2.4m ## Wall Thicknesses - External walls: 0.25m - Internal load-bearing: 0.20m - Internal partitions: 0.10m ## Doors - Main entrance: 1.0m wide x 2.1m tall - Standard internal: 0.9m x 2.1m - Bathroom: 0.8m x 2.1m ## Windows - Standard sill height: 0.9m from floor - Standard window: 1.2m wide x 1.2m tall ## Room Requirements | Room | Min Area | Min Width | |------|----------|-----------| | Bedroom | 10m2 | 2.7m | | Master bedroom | 16m2 | 3.2m | | Kitchen | 8m2 | 2.4m | | Bathroom | 4m2 | 1.5m | | Living room | 20m2 | 3.5m | | Corridor | -- | 1.0m |
When no spec file is found, use these defaults:
| Parameter | Value | |---|---| | Ceiling height (ground/upper) | 2.7 m | | Ceiling height (basement) | 2.4 m | | External wall | 0.25 m | | Partition | 0.10 m | | Main entrance door | 0.9 m x 2.1 m | | Bathroom door | 0.8 m x 2.1 m | | Window sill height | 0.9 m | | Standard window | 1.2 m x 1.2 m | | Bedroom min area | 10 m2 | | Kitchen min area | 8 m2 | | Living room min area | 20 m2 | | Corridor min width | 1.0 m |
A project has BUILDING_SPEC.md with ground floor height 3.5 m (retail), upper floors 2.7 m, external walls 0.30 m, and minimum bedroom area 12 m2.
BUILDING_SPEC.md exists in project rootA model has a bedroom at 9 m2 but the spec requires minimum 10 m2:
No BUILDING_SPEC.md or DIMENSIONS.md in the project:
architectural-dimensions skillarchitectural-dimensions for comprehensive validationBUILDING_SPEC.md template for the client to fill in| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-15 | fail→pass | 19,396 | 2,026 | -90% | 1 | 1 | 0% | 1,176 | 1,867 | +59% | 0 | 0 | — |
case-01 | fail→pass | 5,955 | 15,389 | +158% | 1 | 1 | 0% | 1,032 | 5,124 | +397% | 0 | 0 | — |
case-02 | fail→fail | 9,325 | 13,022 | +40% | 1 | 1 | 0% | 1,630 | 4,325 | +165% | 0 | 0 | — |
case-03 | fail→pass | 22,761 | 15,239 | -33% | 1 | 1 | 0% | 4,926 | 5,056 | +3% | 0 | 0 | — |
case-04 | pass→pass | 13,545 | 17,973 | +33% | 1 | 1 | 0% | 3,009 | 5,749 | +91% | 0 | 0 | — |
case-05 | pass→pass | 13,919 | 11,891 | -15% | 1 | 1 | 0% | 3,202 | 4,461 | +39% | 0 | 0 | — |
case-06 | pass→pass | 17,969 | 21,702 | +21% | 1 | 1 | 0% | 3,077 | 4,707 | +53% | 0 | 0 | — |
case-07 | fail→pass | 5,805 | 2,715 | -53% | 1 | 1 | 0% | 1,140 | 1,934 | +70% | 0 | 0 | — |
case-08 | pass→pass | 15,771 | 1,796 | -89% | 1 | 1 | 0% | 1,564 | 1,814 | +16% | 0 | 0 | — |
case-09 | fail→pass | 7,605 | 2,533 | -67% | 1 | 1 | 0% | 1,321 | 1,990 | +51% | 0 | 0 | — |
case-10 | fail→pass | 10,076 | 2,637 | -74% | 1 | 1 | 0% | 1,801 | 1,951 | +8% | 0 | 0 | — |
case-11 | pass→pass | 7,815 | 2,535 | -68% | 1 | 1 | 0% | 1,566 | 1,955 | +25% | 0 | 0 | — |
case-12 | fail→pass | 10,316 | 3,180 | -69% | 1 | 1 | 0% | 1,779 | 2,119 | +19% | 0 | 0 | — |
case-13 | fail→pass | 9,520 | 2,819 | -70% | 1 | 1 | 0% | 1,843 | 1,960 | +6% | 0 | 0 | — |
case-14 | fail→pass | 13,595 | 2,528 | -81% | 1 | 1 | 0% | 1,174 | 1,945 | +66% | 0 | 0 | — |
case-16 | fail→pass | 7,555 | 1,713 | -77% | 1 | 1 | 0% | 1,316 | 1,822 | +38% | 0 | 0 | — |
case-17 | fail→pass | 8,214 | 2,304 | -72% | 1 | 1 | 0% | 1,392 | 1,872 | +34% | 0 | 0 | — |
case-18 | pass→pass | 10,560 | 4,858 | -54% | 1 | 1 | 0% | 1,995 | 2,430 | +22% | 0 | 0 | — |
case-19 | pass→pass | 11,237 | 4,732 | -58% | 1 | 1 | 0% | 1,955 | 2,444 | +25% | 0 | 0 | — |
case-20 | pass→pass | 8,770 | 3,907 | -55% | 1 | 1 | 0% | 1,388 | 2,150 | +55% | 0 | 0 | — |
case-21 | pass→pass | 16,012 | 4,951 | -69% | 1 | 1 | 0% | 1,383 | 2,445 | +77% | 0 | 0 | — |
case-22 | pass→pass | 7,544 | 2,992 | -60% | 1 | 1 | 0% | 1,497 | 2,192 | +46% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +50 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.