---

name: project-analyze description: This skill should be used when the user asks to "analyze project structure", "explain this codebase", "show dependencies", "find high-risk files", "assess refactor impact", "understand architecture", or run ProjectMind for local codebase analysis. version: 0.1.0 source: fork checksum: f2ceef75cb41ba653bf0b232e2857ccec8bee47a45066cc7cd55e884868b8214 updated_at: 2026-06-08T10:50:00+08:00 layer: research

Codex CLI: Invoke when the description matches, or manually with $project-analyze. No hooks or background auto-run.

Project Intelligence Analysis Skill

When to Use This Skill

Automatically invoke this Skill when:

• User asks "explain this project", "how does this work?"
• User wants to understand architecture or module structure
• User needs impact analysis before refactoring
• User asks about dependencies or file relationships
• User mentions "high-risk files", "critical modules"
• Keywords: "project structure", "architecture", "dependencies", "impact", "understand codebase"

What This Skill Does

ProjectMind Intelligence provides:

1. Knowledge Graph Analysis - Deep project structure understanding
2. Dependency Mapping - Complete relationship visualization
3. Risk Assessment - High-risk file identification
4. Impact Prediction - Change impact estimation
5. Smart Caching - 40s first run, <1s subsequent queries

ProjectMind is a local static-analysis workflow. Do not call external model CLIs, legacy long-context wrappers, hybrid_intelligent_system.py, hybrid_intelligent_system_v2.py, hybrid_intelligent_system_v3.py, or any external AI advisor as part of this skill. Use the local knowledge graph output plus direct repository inspection by Codex.

Instructions

When this Skill is invoked:

Step 1: Extract the Query and Project Path

Identify:

• User's Question: What they want to know
• Project Path: Usually current directory $(pwd) or user-specified path

Step 2: Resolve Runtime and Execute the Local ProjectMind Scan

IMPORTANT: Execute the local ProjectMind knowledge graph scanner before answering architecture, dependency, risk, or impact questions. Prefer python3, fall back to python, and allow PROJECTMIND_HOME to override the default install path.

PROJECT_PATH="[project_path]"
PROJECTMIND_HOME="${PROJECTMIND_HOME:-/Users/WangQiao/claude-enhanced-quality}"
PYTHON_BIN="${PYTHON_BIN:-$(command -v python3 || command -v python || true)}"

if [ -z "$PYTHON_BIN" ]; then
  echo "ProjectMind error: no python3 or python interpreter found" >&2
  exit 1
fi

"$PYTHON_BIN" "$PROJECTMIND_HOME/project_mind.py" "$PROJECT_PATH"

If project path is not specified, use current directory:

PROJECT_PATH="$(pwd)"
PROJECTMIND_HOME="${PROJECTMIND_HOME:-/Users/WangQiao/claude-enhanced-quality}"
PYTHON_BIN="${PYTHON_BIN:-$(command -v python3 || command -v python || true)}"

if [ -z "$PYTHON_BIN" ]; then
  echo "ProjectMind error: no python3 or python interpreter found" >&2
  exit 1
fi

"$PYTHON_BIN" "$PROJECTMIND_HOME/project_mind.py" "$PROJECT_PATH"

Use the scanner output as structured evidence. For query-specific details that the summary does not contain, inspect the repository directly with rg, rg --files, find, sed, and targeted file reads.

Do not use the legacy hybrid intelligent system entrypoints. They delegate to external model tooling and are outside this skill's current execution model.

Step 3: Present Results

Format the output as:

## 🧠 Project Intelligence Analysis

**Query**: [User's Question]
**Project**: [Project Path]
**Analysis Time**: [Cached/40s] ⚡

### 📊 Project Overview

- **Total Files**: XXX
- **Lines of Code**: XXX,XXX
- **Main Technologies**: [React, TypeScript, Node.js, etc.]
- **Architecture Pattern**: [MVC, Microservices, Monolith, etc.]

### 🎯 Answer to Your Question

[Direct answer to the user's specific question based on ProjectMind output and targeted local code inspection]

### 🏗️ Architecture Insights

**Module Structure**:

Core Modules: ├── [Module 1] (XX files, core business logic) ├── [Module 2] (XX files, data layer) └── [Module 3] (XX files, API layer)

Supporting Modules: ├── [Utils] (XX files) └── [Components] (XX files)

**Key Relationships**:
- [Module A] depends heavily on [Module B]
- [File X] is central hub (XX dependencies)
- [Component Y] is isolated (low coupling)

### ⚠️ High-Risk Areas

| File/Module | Risk Level | Reason | Dependencies |
|-------------|------------|--------|--------------|
| [Path 1] | 🔴 Critical | Core auth logic, 45 dependencies | 45 files |
| [Path 2] | 🟠 High | Payment processing | 32 files |
| [Path 3] | 🟡 Medium | Complex business rules | 18 files |

### 💡 Recommendations

**If Planning Changes**:
1. **[Specific Module]**
   - Files Affected: XX files
   - Risk Level: [Low/Medium/High]
   - Test Requirements: [Unit/Integration/E2E]
   - Estimated Effort: X hours

2. **[Another Module]**
   - Impact: [Detailed impact analysis]
   - Mitigation: [How to reduce risk]

**Architecture Improvements**:
- [Specific decoupling opportunity]
- [Specific pattern improvement]
- [Specific technical debt reduction]

### 🔍 Dependency Visualization

```mermaid
graph TD
    A[Core Module] --> B[Feature 1]
    A --> C[Feature 2]
    B --> D[Util 1]
    C --> D
    B --> E[DB Layer]
    C --> E

(Note: Mermaid diagram representing key dependencies)

🚨 Change Impact Assessment

Before modifying [specific file/module]:

• ✅ Test these files: [List]
• ⚠️ Watch for side effects in: [List]
• 🔍 Review integration points: [List]
• ⏱️ Estimated impact: [Low/Medium/High]

📈 Technical Metrics

• Modularity Score: XX/100
• Coupling Level: [Low/Medium/High]
• Code Churn: [Files frequently changed]
• Critical Path: [Most important business logic files]

### Step 4: Provide Actionable Guidance

Offer specific next steps:
- Code locations to examine
- Tests to run before changes
- Files to backup before refactoring
- Team members to consult (if AI annotations present)

## Examples

### Example 1: Understanding Architecture
**User**: "Explain how the authentication system works"

**Execute**:
```bash
PROJECT_PATH="$(pwd)"
PROJECTMIND_HOME="${PROJECTMIND_HOME:-/Users/WangQiao/claude-enhanced-quality}"
PYTHON_BIN="${PYTHON_BIN:-$(command -v python3 || command -v python || true)}"
"$PYTHON_BIN" "$PROJECTMIND_HOME/project_mind.py" "$PROJECT_PATH"

Present: Architecture explanation with module relationships, file paths, and authentication flow diagram, using ProjectMind output plus targeted rg searches for authentication entry points.

Example 2: Refactoring Impact

User: "I want to refactor the database layer, what's the impact?"

Execute:

PROJECT_PATH="$(pwd)"
PROJECTMIND_HOME="${PROJECTMIND_HOME:-/Users/WangQiao/claude-enhanced-quality}"
PYTHON_BIN="${PYTHON_BIN:-$(command -v python3 || command -v python || true)}"
"$PYTHON_BIN" "$PROJECTMIND_HOME/project_mind.py" "$PROJECT_PATH"

Present: Impact analysis showing affected files, risk assessment, and refactoring strategy, verified with local dependency/file searches.

Example 3: Finding High-Risk Code

User: "What are the most critical files in this project?"

Execute:

PROJECT_PATH="$(pwd)"
PROJECTMIND_HOME="${PROJECTMIND_HOME:-/Users/WangQiao/claude-enhanced-quality}"
PYTHON_BIN="${PYTHON_BIN:-$(command -v python3 || command -v python || true)}"
"$PYTHON_BIN" "$PROJECTMIND_HOME/project_mind.py" "$PROJECT_PATH"

Present: List of high-risk files with dependency counts, complexity scores, and business criticality.

Performance Characteristics

• First Analysis: ~40 seconds (deep project scan)
• Cached Queries: <1 second (instant response)
• File Limit: 300 files deep analysis
• Cache Invalidation: Automatic on file changes

Smart Caching System

The system automatically caches:

• ✅ Project structure analysis
• ✅ Dependency mappings
• ✅ Code entity relationships
• ✅ Risk assessments

Cache updates when:

• Files are modified
• Dependencies change
• Project structure updates

Integration with Other Systems

Git Memory Integration

• Tracks commit intentions automatically
• Analyzes team collaboration patterns
• Detects potentially conflicting changes

CodeDNA Integration

• Project-level quality scoring
• Hotspot identification with quality metrics
• ROI-optimized refactoring prioritization

AI-Specific Annotations

Recognizes and utilizes:

/*@ai:risk=1-5|deps=file1,file2|core=true|chain=auth*/

Annotation fields:

• risk: 1 (safe) to 5 (critical)
• deps: Key dependencies
• core: Core functionality flag
• chain: Business logic chain identifier
• api: API type (internal/external)
• auth: Authentication requirements

Important Notes

• Always execute the Python command, don't simulate analysis
• Never use external model CLIs or the legacy hybrid AI analysis scripts for this skill
• Use current directory as default project path
• Explain relationships, not just list files
• Provide visual diagrams when helpful
• Assess real impact, not theoretical concerns
• Offer specific actions, not generic advice

Common Analysis Queries

Architecture Understanding

• "How does [feature] work?"
• "What's the data flow for [process]?"
• "Explain the module structure"

Impact Analysis

• "Impact of changing [file/module]?"
• "What depends on [component]?"
• "Risk of refactoring [area]?"

Code Navigation

• "Where is [functionality] implemented?"
• "Which files handle [feature]?"
• "What are the entry points?"

Quality Assessment

• "What are the problem areas?"
• "Where should we focus refactoring?"
• "What's the technical debt?"

Prerequisites

• Python 3 preferred, or Python fallback
• ProjectMind V2 system installed at /Users/WangQiao/claude-enhanced-quality or another path supplied through PROJECTMIND_HOME
• Read access to project files
• Write access for caching (automatic)

Fallback Strategy

DO NOT use legacy hybrid scripts (hybrid_intelligent_system.py, hybrid_intelligent_system_v2.py, or hybrid_intelligent_system_v3.py) - they delegate to external model tooling and are deprecated for this skill.

If the local ProjectMind scan fails:

1. Check project path is correct
2. Verify python3 or python is available with command -v python3 || command -v python
3. Verify $PROJECTMIND_HOME/project_mind.py exists, defaulting to /Users/WangQiao/claude-enhanced-quality/project_mind.py
4. Examine error message
5. Fall back to direct local inspection with rg, rg --files, and targeted file reads
6. Report the issue and mark which parts are based on fallback inspection

Performance Tips

• First query takes 40s to build knowledge graph
• Subsequent queries are instant (cached)
• Narrow queries get faster responses
• Broader queries provide more context