🏗️ Construction Project Management Support Agent Skill

A production-grade AI agent skill that evaluates construction project plans for schedule, cost control, and site risk management against PMI/Lean Construction standards.

---

📋 Overview

The Construction Project Management Support skill is a sophisticated AI agent that provides professional-grade project management assessments grounded in world-renowned frameworks. It conducts comprehensive evaluations across five critical dimensions:

• Schedule & CPM Integrity (25%) — Network logic, float analysis, critical path realism
• Cost Control & EVM (25%) — Budget structure, CPI/SPI analysis, contingency adequacy
• Risk & Safety Management (20%) — Risk register completeness, mitigation planning, safety protocols
• Stakeholder & Procurement (15%) — RACI clarity, contract terms, supply chain resilience
• Quality & Change Control (15%) — QA/QC processes, change order discipline, inspection regimes

---

✨ Key Features

🔬 Evidence-Based Assessment

Every score and recommendation is backed by authoritative sources:

• PMI PMBOK Guide (7th Edition) — Global project management standard
• Earned Value Management (EIA-748) — Quantitative cost/schedule analysis
• Lean Construction & Last Planner System — Workflow optimization and pull planning
• Critical Path Method (AACE RP 46R-11) — Network scheduling best practices
• ISO 21500 & RICS NRM — International guidance and measurement standards

🎯 Multi-Dimensional Scoring

Comprehensive 0-100 scoring rubric with weighted dimensions, evidence citations, and letter grades (A-F).

🗺️ Prioritized Improvement Roadmap

Effort/impact-ranked recommendations with implementation phases, timelines, and expected outcomes.

🧠 Self-Improving Knowledge Base

Automated weekly knowledge updates via crawl pipeline, keeping the skill current with latest research and standards.

🔄 Graceful Degradation

Works offline using cached knowledge when web search unavailable—no single point of failure.

---

🚀 Use Cases

| Scenario | Description | Output | |----------|-------------|--------| | Schedule Health | "Is my project schedule realistic?" | CPM analysis, duration validation, recovery roadmap | | EVM Diagnosis | "CPI is 0.85, SPI is 0.9 — what do I do?" | Variance analysis, root cause identification, corrective actions | | Risk Review | "Review our risk register" | Completeness check, gap analysis, mitigation enhancement | | Stakeholder Mapping | "Who signs off at each stage?" | RACI matrix, approval chains, communication plan | | Comprehensive Audit | Full project health assessment | Multi-dimensional scores with prioritized roadmap |

---

📁 Project Structure

construction-project-management-support/
├── README.md                                  # This file
├── LICENSE                                     # MIT License
├── CLAUDE.md                                   # Project instructions
├── PROJECT-detail.md                           # Technical specification
├── PROJECT-DEVELOPMENT-PHASE-TRACKING.md       # Development status
├── SECOND-KNOWLEDGE-BRAIN.md                    # Knowledge base (self-updating)
│
├── skills/                                     # Agent skill definitions
│   ├── main.md                                 # Main harness (8-stage workflow)
│   ├── sub-intake.md                           # Intake & context gathering
│   ├── sub-framework-selector.md               # Framework selection logic
│   ├── sub-scoring-engine.md                   # Multi-dimensional scoring
│   └── sub-improvement-roadmap.md              # Prioritized recommendations
│
├── tools/                                      # Supporting tools
│   └── knowledge_updater.py                    # Weekly knowledge crawl pipeline
│
└── tests/                                      # Test scenarios
    └── test-scenarios.md                       # 5 comprehensive test cases

---

🏗️ Harness Architecture

The skill implements a rigorous 8-stage workflow:

┌─────────────────────────────────────────────────────────────────┐
│                     1. INTAKE (sub-intake)                     │
│  Gather structured inputs, scope, goals, constraints             │
└───────────────────────────┬─────────────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────────────┐
│              2. FRAMEWORK SELECTOR (sub-framework)              │
│  Select PMI/EVM/Lean/CPM/ISO frameworks with justification    │
└───────────────────────────┬─────────────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────────────┐
│                      3. RESEARCH                                 │
│  Gather evidence from WebSearch + SECOND-KNOWLEDGE-BRAIN        │
└───────────────────────────┬─────────────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────────────┐
│                 4. SCORING ENGINE (sub-scoring)                 │
│  Apply 0-100 rubric with evidence citations, weighted scores   │
└───────────────────────────┬─────────────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────────────┐
│                      5. CHALLENGE                                │
│  Devil's advocate review: test assumptions, grade certainty     │
└───────────────────────────┬─────────────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────────────┐
│            6. IMPROVEMENT ROADMAP (sub-roadmap)                 │
│  Prioritized effort/impact recommendations with phases          │
└───────────────────────────┬─────────────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────────────┐
│                     7. SYNTHESIZE                               │
│  Assemble professional report with all sections                │
└───────────────────────────┬─────────────────────────────────────┘
                            │
┌───────────────────────────▼─────────────────────────────────────┐
│                   8. QUALITY GATES                              │
│  Verify citations, roadmap traceability, limitations stated     │
└─────────────────────────────────────────────────────────────────┘

---

📊 Scoring Rubric

Each dimension is scored 0-100 with specific criteria:

| Score Range | Grade | Description | |-------------|-------|-------------| | 90-100 | A | Excellent — Industry best practice, comprehensive coverage, proactive management | | 75-89 | B | Good — Solid practice, minor gaps, generally well-managed | | 60-74 | C | Fair — Basic practice in place, notable gaps, reactive approach | | 40-59 | D | Poor — Significant gaps, minimal formal process, high exposure | | 0-39 | F | Critical — Major deficiencies, unacceptable risk, immediate attention required |

Overall Grade Calculation:

Overall = (Schedule × 0.25) + (Cost × 0.25) + (Risk × 0.20) +
          (Stakeholder × 0.15) + (Quality × 0.15)

---

🧠 Knowledge Base

The skill maintains a comprehensive knowledge base (SECOND-KNOWLEDGE-BRAIN.md) containing:

• Framework Documentation — Detailed coverage of PMI, EVM, Lean, CPM, ISO, RICS
• Research Papers — Key studies with relevance notes and citations
• Best Practices — State-of-the-art methods and industry tools
• Authoritative Sources — PMI, LCI, RICS, ISO, CII, AACE, and more

Self-Update Pipeline

The tools/knowledge_updater.py script runs weekly (recommended via cron) to:

1. Fetch latest research from authoritative sources
2. Parse and score entries by relevance
3. Deduplicate against existing entries
4. Append new findings to knowledge base

Install dependencies (optional):

pip install crawl4ai

Run manually:

python tools/knowledge_updater.py --dry-run --verbose

Set up weekly cron:

0 2 * * 0 cd /path/to/construction-project-management-support && python tools/knowledge_updater.py >> tools/knowledge_updater.log 2>&1

---

🧪 Testing

The skill includes 5 comprehensive test scenarios in tests/test-scenarios.md:

1. Schedule Health — Commercial building, planning phase
2. EVM Diagnosis — Infrastructure project, execution phase
3. Risk Register Review — Healthcare construction, regulatory focus
4. Stakeholder Mapping — School project, public sector approvals
5. Degraded Mode — Offline operation using cached knowledge

Each scenario documents expected outputs stage-by-stage with pass criteria.

---

🛠️ Installation

For Claude Code Users

1. Clone the repository:

git clone https://github.com/dungnotnull/construction-project-management-support-agent-skill.git
cd construction-project-management-support-agent-skill

2. Verify skill structure:

# Should show all skill files
ls -la skills/

3. Test the skill:

• Open Claude Code in the project directory
• Invoke the skill with a construction project management query
• Example: "Review this project schedule for realism"

For Integration

The skill follows the Claude Code skill format with frontmatter metadata:

---
name: construction-project-management-support
description: Evaluates constructio