メニュー
Compare IFC model data against PDF drawing schedules. Reads saved extraction data from output/, identifies count mismatches, missing properties, and conflicting values. Generates Markdown discrepancy report with PASS/WARN/INFO flags.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Reference manual for BuildBrain CLI tools (ifc_extract.py and pdf_extract.py). Shows all commands, flags, output formats, and type shorthands. Use this as a reference when working with IFC or PDF data.
Extract data from IFC building models. Accepts element types (door, wall, window, slab, roof, etc.) and optional property filters. Translates natural language queries into CLI commands. Results saved to output/ for cross-validation.
Extract schedules, text, and tables from PDF construction drawings. Can search for specific terms to find relevant pages. Results saved to output/ for cross-validation.
Generate formatted construction deliverables — QTO tables, element schedules, material takeoffs, compliance checklists. Reads extracted data from output/, uses quantities command for aggregates. Formats per Australian construction conventions.
Scan IFC and PDF files to create a project inventory. Shows element counts, schedule locations, and data quality issues. Use when starting work on a new project or wanting an overview.
| name | cross-validate |
| description | Compare IFC model data against PDF drawing schedules. Reads saved extraction data from output/, identifies count mismatches, missing properties, and conflicting values. Generates Markdown discrepancy report with PASS/WARN/INFO flags. |
| allowed-tools | Bash, Read, Write, Grep |
| argument-hint | [element-type] |
Compare IFC model data against PDF schedules and generate a discrepancy report.
Check output/ for previously saved extraction files:
output/ifc_*.json (prefer type-specific files)output/pdf_schedules_*.json (prefer over generic pdf_*.json)Use the most recent file for each. If no data found, tell user to run /ifc-extract and /pdf-extract first.
Avoid context overflow. If user didn't specify a type, pick the first type with data in both sources.
Normalise names: "D01" ↔ "Door 01" ↔ "D-01". Strip prefixes, separators, leading zeros. Case-insensitive.
| State | Meaning | Example |
|---|---|---|
| MISMATCH | IFC says X, PDF says Y | IFC: FRL-30, PDF: FRL-60 |
| ABSENT | IFC property null, PDF has value | IFC: FireRating null, PDF: FRL-60 |
| PASS | Both sources agree | IFC: 47 doors, PDF: 47 rows |
Likely cause: IFC export settings — "Export IFC common property sets" may not be enabled. Action: Architect should re-export with common property sets enabled.
Use this format:
## Discrepancy Report — <Project Name> — <YYYY-MM-DD>
### WARN: <Title>
- IFC Model: <value>
- PDF <Schedule> (p.<page>): <value>
- Action: <recommendation>
### PASS: <Title>
- IFC Model: <value>
- PDF <Schedule> (p.<page>): <value>
Order: WARN (MISMATCH) → WARN (ABSENT) → INFO → PASS
End with summary: total checks, PASS count, WARN count, INFO count.
Save to output/discrepancy_<YYYY-MM-DD_HHMMSS>.md.
count field or array length from JSON data./ifc-extract and /pdf-extract.