datadrivenconstruction/DDC_Skills_for_AI_Agents_in_Construction

DDC Skills Collection for AI Coding Assistants

What it is – A curated set of 238 “skills” (each a SKILL.md file with problem description, code template and required tools) that let an AI coding assistant (Claude Code, Cursor, Copilot, etc.) generate ready‑to‑run scripts for common construction‑company workflows.

Why it matters – Construction firms have a lot of data (Excel estimates, BIM/IFC models, PDFs, site photos) that is usually siloed and manually processed. These skills turn that data into automated pipelines, dashboards, and AI‑driven analyses without the user having to write the boiler‑plate code themselves.

Main categories

Category	Typical content	When to start
1 _DDC_Toolkit	Production‑ready tools: CWICR cost database, CAD‑to‑Excel converters, KPI dashboards	Need a working tool immediately
2 _DDC_Book	Skills mapped to chapters of the free Data‑Driven Construction book (data maturity, estimation, ML)	Want a structured learning path
3 _DDC_Insights	n8n workflow templates, AI‑agent orchestration, field‑report bots	Looking for quick workflow automation
4 _DDC_Curated	Document‑generation (PDF/Excel/DOCX/PPTX) and quality‑check scripts	Need report or document templates
5 _DDC_Innovative	Computer‑vision defect detection, digital twins, ESG & EU‑AI‑Act checks, generative design	Ready to experiment with AI/ML
6 _OpenConstructionERP	Scripts that plug into the open‑source OpenConstructionERP platform (cost bases, BOQ, 4D/5D scheduling, tendering)	Building on an existing ERP

Typical workflow

1. Audit – Run data-silo-detection to list all data sources.
2. Classify – Use data-type-classifier to label structured, semi‑structured or unstructured data.
3. Connect – Deploy an etl-pipeline or one of the CAD converters (ifc-to-excel, rvt-to-excel).
4. Automate – Set up recurring reports with n8n-daily-report or a KPI dashboard.
5. Analyze – Apply ML skills such as cost-prediction or schedule-forecaster.

How to use a skill

# Clone the repo
git clone https://github.com/datadrivenconstruction/DDC_Skills_for_AI_Agents_in_Construction.git

# Open a skill folder in your AI assistant (e.g. Claude Code)
cd DDC_Skills_for_AI_Agents_in_Construction/1_DDC_Toolkit/CWICR-Database/semantic-search-cwicr/

# The assistant reads SKILL.md, generates the code, you review/adapt it, then run it.

A minimal example (etl-pipeline) is provided in the README and shows a simple pandas‑based extract‑transform‑load script.

Prerequisites

• Python 3.9+ (most skills are Python scripts)
• An AI coding assistant that can read SKILL.md and emit code
• Basic Python knowledge and access to your own data files
• Optional: Docker (for n8n workflows), a PostgreSQL/SQLite DB, OpenAI API key for LLM‑based skills

Installation of common dependencies

pip install pandas openpyxl ifcopenshell pdfplumber

Who should look at this

• Construction executives / ops managers – the GETTING_STARTED.md gives a non‑technical overview and a prioritized list of pain points.
• Developers / IT leads – can pick any skill, let the AI generate the code, then integrate it into existing systems.
• Researchers – the “Innovative” folder contains experimental AI/ML pipelines (defect detection, digital twins) that can be extended.

Documentation & support

• GETTING_STARTED.md – high‑level onboarding
• OPTIMIZER_GUIDE.md – tips for getting the most out of AI assistants
• IMPROVEMENT_ROADMAP.md – how contributors can add new skills
• Free multilingual book Data‑Driven Construction (31 languages) explains the methodology behind every skill.

License – MIT for the tools and skills; the CWICR database is CC BY‑NC 4.0 (non‑commercial) with a separate commercial license.

---

Bottom line – The repository is a practical toolbox that bridges AI coding assistants and the messy, data‑heavy world of construction, letting companies automate estimation, reporting, BIM data extraction, and even advanced AI analyses with minimal hand‑coding.