Skip to main content
skill CX Codex

specification-extractor — Claude Code skill: install

The specification-extractor skill turns construction specification PDFs into structured sections, product requirements, standards, and submittal data for estimating and procurement.

MIT 128 installs ★ 434 updated 2026-10-19

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add datadrivenconstruction/DDC_Skills_for_AI_Agents_in_Construction --skill specification-extractor -a claude-code
Open on GitHub

What specification-extractor does

specification-extractor reads construction specification PDFs and pulls out CSI MasterFormat sections, requirements, product details, referenced standards, and required submittals. Its structured result covers document sections, manufacturers, product names, models, properties, submittal details, and page totals.

The implementation uses Python with pdfplumber to read page text, detect section and part headings, and organize the extracted content. It separates Part 1 general information, Part 2 products, and Part 3 execution. The skill requires python3; without it, extraction cannot run.

What it helps with

  • parse a construction specification PDF
  • identify CSI MasterFormat sections
  • extract product requirements and standards
  • list required submittals
  • prepare specification data for estimating and procurement

Example prompts for the agent

  • Read the attached construction specification PDF and extract its CSI sections, requirements, and submittals.
  • Find the manufacturers, product names, models, properties, and standards referenced in this specification.
  • Create a structured list of every required submittal, including its type, description, timing, and copy count.

How to install specification-extractor

1Skills CLI (Vercel)

npx skills add datadrivenconstruction/DDC_Skills_for_AI_Agents_in_Construction --skill specification-extractor -a claude-code

2Manually: put the files in place

~/.claude/skills/specification-extractor/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/specification-extractor/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about specification-extractor

How do I install specification-extractor in Claude Code?

Run npx skills add datadrivenconstruction/DDC_Skills_for_AI_Agents_in_Construction --skill specification-extractor -a claude-code, then check ~/.claude/skills/specification-extractor/SKILL.md.

What does specification-extractor do?

It extracts CSI sections, requirements, product information, referenced standards, and submittals from construction specification PDFs, then organizes them as structured data.

What does specification-extractor need to run?

The skill requires python3. Its PDF text extraction uses pdfplumber.

Source text

SKILL.md · MIT
---
name: "specification-extractor"
description: "Extract structured data from construction specifications. Parse CSI sections, requirements, submittals, and product data from spec documents."
homepage: "https://datadrivenconstruction.io"
metadata: {"openclaw": {"emoji": "📑", "os": ["darwin", "linux", "win32"], "homepage": "https://datadrivenconstruction.io", "requires": {"bins": ["python3"]}}}
---
# Specification Extractor for Construction

## Overview

Extract structured data from construction specification documents. Parse CSI MasterFormat sections, identify requirements, submittals, product standards, and compile actionable data for estimating and procurement.

## Business Case

Automated spec extraction enables:
- **Faster Estimating**: Quickly identify scope and requirements
- **Procurement Accuracy**: Extract exact product specifications
- **Submittal Tracking**: Identify all required submittals
- **Compliance Checking**: Verify specs against standards

## Technical Implementation

```python
from dataclasses import dataclass, field
from typing import List, Dict, Any, Optional
import re
import pdfplumber
from pathlib import Path

@dataclass
class SpecSection:
    number: str  # e.g., "03 30 00"
    title: str
    part1_general: Dict[str, Any]
    part2_products: Dict[str, Any]
    part3_execution: Dict[str, Any]
    raw_text: str

@dataclass
class ProductRequirement:
    section: str
    manufacturer: str
    product_name: str
    model: str
    standards: List[str]
    properties: Dict[str, str]

@dataclass
class SubmittalRequirement:
    section: str
    submittal_type: str  # shop drawings, samples, product data, etc.
    description: str
    timing: str
    copies: int

@dataclass
class SpecExtractionResult:
    document_name: str
    total_pages: int
    sections: List[SpecSection]
    products: List[ProductRequirement]
    submittals: List[SubmittalRequirement]
    standards_referenced: List[str]

class SpecificationExtractor:
    """Extract structured data from construction specifications."""

    # CSI MasterFormat patterns
    CSI_SECTION_PATTERN = r'^(\d{2}\s?\d{2}\s?\d{2})\s*[-–]\s*(.+?)$'
    PART_PATTERN = r'^PART\s+(\d+)\s*[-–]\s*(.+?)$'
    ARTICLE_PATTERN = r'^(\d+\.\d+)\s+([A-Z][A-Z\s]+)$'

    # Submittal type keywords
    SUBMITTAL_TYPES = {
        'shop drawings': 'Shop Drawings',
        'product data': 'Product Data',
        'samples': 'Samples',
        'certificates': 'Certificates',
        'test reports': 'Test Reports',
        'manufacturer instructions': 'Manufacturer Instructions',
        'warranty': 'Warranty',
        'maintenance data': 'Maintenance Data',
        'mock-ups': 'Mock-ups',
    }

    # Common standard organizations
    STANDARD_PATTERNS = [
        r'ASTM\s+[A-Z]\d+',
        r'ANSI\s+[A-Z]?\d+',
        r'ACI\s+\d+',
        r'AISC\s+\d+',
        r'AWS\s+[A-Z]\d+',
        r'ASCE\s+\d+',
        r'UL\s+\d+',
        r'FM\s+\d+',
        r'NFPA\s+\d+',
        r'IBC\s+\d+',
    ]

    def __init__(self):
        self.sections: Dict[str, SpecSection] = {}

    def extract_from_pdf(self, pdf_path: str) -> SpecExtractionResult:
        """Extract specification data from PDF."""
        path = Path(pdf_path)

        all_text = ""
        page_count = 0

        with pdfplumber.open(pdf_path) as pdf:
            page_count = len(pdf.pages)
            for page in pdf.pages:
                text = page.extract_text() or ""
                all_text += text + "\n\n"

        # Parse sections
        sections = self._parse_sections(all_text)

        # Extract products
        products = self._extract_products(sections)

        # Extract submittals
        submittals = self._extract_submittals(sections)

        # Extract standards
        standards = self._extract_standards(all_text)

        return SpecExtractionResult(
            document_name=path.name,
            total_pages=page_count,
            sections=sections,
            products=products,
            submittals=submittals,
            standards_referenced=standards
        )

    def _parse_sections(self, text: str) -> List[SpecSection]:
        """Parse CSI sections from specification text."""
        sections = []
        lines = text.split('\n')

        current_section = None
        current_part = None
        current_content = []

        for line in lines:
            line = line.strip()
            if not line:
                continue

            # Check for section header
            section_match = re.match(self.CSI_SECTION_PATTERN, line, re.IGNORECASE)
            if section_match:
                # Save previous section
                if current_section:
                    sections.append(self._finalize_section(current_section, current_content))

                current_section = {
                    'number': section_match.group(1).replace(' ', ''),
                    'title': section_match.group(2).strip(),
                    'parts': {}
                }
                current_content = []
                current_part = None
                continue

            # Check for part header
            part_match = re.match(self.PART_PATTERN, line, re.IGNORECASE)
            if part_match and current_section:
                part_num = part_match.group(1)
                part_name = part_match.group(2).strip()
                current_part = f"part{part_num}"
                current_section['parts'][current_part] = {
                    'name': part_name,
                    'content': []
                }
                continue

            # Add content to current part
            if current_section and current_part:
                current_section['parts'][current_part]['content'].append(line)
            elif current_section:
                current_content.append(line)

        # Save last section
        if current_section:
            sections.append(self._finalize_section(current_section, current_content))

        return sections

    def _finalize_section(self, section_data: Dict, general_content: List[str]) -> SpecSection:
        """Finalize a section with parsed parts."""
        parts = section_data.get('parts', {})

        part1 = self._parse_part_content(parts.get('part1', {}).get('content', []))
        part2 = self._parse_part_content(parts.get('part2', {}).get('content', []))
        part3 = self._parse_part_content(parts.get('part3', {}).get('content', []))

        return SpecSection(
            number=section_data['number'],
            title=section_data['title'],
            part1_general=part1,
            part2_products=part2,
            part3_execution=part3,
            raw_text='\n'.join(general_content)
        )

    def _parse_part_content(self, content: List[str]) -> Dict[str, Any]:
        """Parse part content into structured data."""
        result = {
            'articles': {},
            'items': []
        }

        current_article = None

        for line in content:
            # Check for article header
            article_match = re.match(self.ARTICLE_PATTERN, line)
            if article_match:
                current_article = article_match.group(1)
                result['articles'][current_article] = {
                    'title': article_match.group(2),
                    'items': []
                }
                continue

            # Add to current article or general items
            if current_article and current_article in result['articles']:
                result['articles'][current_article]['items'].append(line)
            else:
                result['items'].append(line)

        return result

    def _extract_products(self, sections: List[SpecSection]) -> List[ProductRequirement]:
        """Extract product requirements from Part 2."""
        products = []

        for section in sections:
            part2 = section.part2_products

            for article_num, article in part2.get('articles', {}).items():
                if 'MANUFACTURERS' in article['title'].upper():
                    for item in article['items']:
                        # Extract manufacturer names
                        if item.strip().startswith(('A.', 'B.', 'C.', '1.', '2.', '3.')):
                            mfr_name = re.sub(r'^[A-Z\d]+\.\s*', '', item).strip()
                            products.append(ProductRequirement(
                                section=section.number,
                                manufacturer=mfr_name,
                                product_name='',
                                model='',
                                standards=[],
                                properties={}
                            ))

                elif 'MATERIALS' in article['title'].upper() or 'PRODUCTS' in article['title'].upper():
                    for item in article['items']:
                        # Extract material requirements
                        standards = self._extract_standards(item)
                        if standards:
                            products.append(ProductRequirement(
                                section=section.number,
                                manufacturer='',
                                product_name=item[:100],
                                model='',
                                standards=standards,
                                properties={}
                            ))

        return products

    def _extract_submittals(self, sections: List[SpecSection]) -> List[SubmittalRequirement]:
        """Extract submittal requirements from Part 1."""
        submittals = []

        for section in sections:
            part1 = section.part1_general

            for article_num, article in part1.get('articles', {}).items():
                if 'SUBMITTAL' in article['title'].upper():
                    for item in article['items']:
                        item_lower = item.lower()

                        for keyword, submittal_type in self.SUBMITTAL_TYPES.items():
                            if keyword in item_lower:
                                submittals.append(SubmittalRequirement(
                                    section=section.number,
                                    submittal_type=submittal_type,
                                    description=item.strip(),
                                    timing='Prior to fabrication',
                                    copies=3
                                ))
                                break

        return submittals

    def _extract_standards(self, text: str) -> List[str]:
        """Extract referenced standards from text."""
        standards = []

        for pattern in self.STANDARD_PATTERNS:
            matches = re.findall(pattern, text, re.IGNORECASE)
            standards.extend(matches)

        return list(set(standards))

    def generate_submittal_log(self, result: SpecExtractionResult) -> str:
        """Generate submittal log from extraction results."""
        lines = ["# Submittal Log", ""]
        lines.append(f"**Project Specs:** {result.document_name}")
        lines.append(f"**Total Submittals:** {len(result.submittals)}")
        lines.append("")

        lines.append("| # | Section | Type | Description | Status |")
        lines.append("|---|---------|------|-------------|--------|")

        for i, sub in enumerate(result.submittals, 1):
            desc = sub.description[:50] + "..." if len(sub.description) > 50 else sub.description
            lines.append(f"| {i} | {sub.section} | {sub.submittal_type} | {desc} | Pending |")

        return "\n".join(lines)

    def generate_product_schedule(self, result: SpecExtractionResult) -> str:
        """Generate product schedule from extraction results."""
        lines = ["# Product Schedule", ""]

        # Group by section
        by_section = {}
        for prod in result.products:
            if prod.section not in by_section:
                by_section[prod.section] = []
            by_section[prod.section].append(prod)

        for section, products in sorted(by_section.items()):
            lines.append(f"## Section {section}")
            lines.append("")

            for prod in products:
                if prod.manufacturer:
                    lines.append(f"- **Manufacturer:** {prod.manufacturer}")
                if prod.product_name:
                    lines.append(f"- **Product:** {prod.product_name}")
                if prod.standards:
                    lines.append(f"- **Standards:** {', '.join(prod.standards)}")
                lines.append("")

        return "\n".join(lines)

    def generate_report(self, result: SpecExtractionResult) -> str:
        """Generate comprehensive extraction report."""
        lines = ["# Specification Extraction Report", ""]
        lines.append(f"**Document:** {result.document_name}")
        lines.append(f"**Pages:** {result.total_pages}")
        lines.append(f"**Sections Found:** {len(result.sections)}")
        lines.append("")

        # Sections summary
        lines.append("## Sections Extracted")
        for section in result.sections:
            lines.append(f"- **{section.number}** - {section.title}")
        lines.append("")

        # Standards
        if result.standards_referenced:
            lines.append("## Standards Referenced")
            for std in sorted(set(result.standards_referenced)):
                lines.append(f"- {std}")
            lines.append("")

        # Submittals summary
        lines.append("## Submittals Required")
        lines.append(f"Total: {len(result.submittals)}")
        by_type = {}
        for sub in result.submittals:
            by_type[sub.submittal_type] = by_type.get(sub.submittal_type, 0) + 1
        for t, count in sorted(by_type.items()):
            lines.append(f"- {t}: {count}")
        lines.append("")

        # Products summary
        lines.append("## Products/Manufacturers")
        lines.append(f"Total: {len(result.products)}")

        return "\n".join(lines)
```

## Quick Start

```python
# Initialize extractor
extractor = SpecificationExtractor()

# Extract from PDF
result = extractor.extract_from_pdf("Project_Specifications.pdf")

print(f"Found {len(result.sections)} sections")
print(f"Found {len(result.submittals)} submittals")
print(f"Found {len(result.products)} product requirements")

# Generate submittal log
submittal_log = extractor.generate_submittal_log(result)
print(submittal_log)

# Generate product schedule
product_schedule = extractor.generate_product_schedule(result)
print(product_schedule)

# Full report
report = extractor.generate_report(result)
print(report)
```

## Dependencies

```bash
pip install pdfplumber
```

© datadrivenconstruction · MIT · datadrivenconstruction/DDC_Skills_for_AI_Agents_in_Construction

Similar skills

All skills for Data and databases →

From the same set156

The whole set →
safety-compliance-checker safety-compliance-checker is a Claude Code skill that checks safety compliance on construction sites. It tracks PPE, zone access, and work at heights, and generates violation reports. CX Codex +1 131installs daily-report-generator The daily-report-generator skill pulls weather, crew hours, and progress data from your project sources and assembles them into a polished PDF report. Cuts manual reporting time from 45–60 minutes to about 9 minutes for review. CX Codex +1 130installs energy-simulation The energy-simulation skill models building energy performance, from heating and cooling loads to envelope checks and HVAC sizing. Use it when reviewing energy-code scenarios. CX Codex +1 129installs sensor-data-aggregator The sensor-data-aggregator skill collects, aggregates, and analyzes construction-site IoT sensor data in real time, detecting anomalies and triggering environmental alerts. CX Codex +1 129installs incident-reporting incident-reporting logs construction site incidents, conducts 5 Whys investigations, and tracks corrective actions. Requires Python 3. CX Codex +1 128installs contractor-matching-ai contractor-matching-ai matches construction project requirements with contractor profiles and ranks suitable candidates. Use it for an initial selection based on several criteria. CX Codex +1 128installs cost-prediction cost-prediction forecasts construction project budgets using historical data with Linear Regression, KNN, and Random Forest models. Requires Python 3 and scikit-learn to train on your project dataset. CX Codex +1 128installs docx-construction docx-construction generates construction Word documents such as subcontracts, bid proposals, daily jobsite logs, and transmittal letters using dynamic Python templates. CX Codex +1 128installs

Project sponsors

Companies that keep this analytics open