vm-deterministic-extractor

Skill detail with category, linked agents, and source metadata.

vm-deterministic-extractor

Deterministic script-based extraction skill for reverse engineering. Generates and executes OS-aware scripts (PowerShell on Windows, POSIX on Linux/macOS) to extract file inventories, signatures, dependency graphs, LOC counts, and complexity metrics WITHOUT using AI. Reads script-config.json from Phase -1 calibration to tune regex patterns. All structural extraction is deterministic and reproducible. USE WHEN: running Phases 0-4 of reverse engineering where structural data can be extracted by scripts rather than AI.

Category: Code Generation Used by 1 agents

Source: .github/skills/code-generation/vm-deterministic-extractor/SKILL.md

Used By Agents

3 - Analysis - Reverse Engineering Agent

Preview

View source preview (first 3000 chars)


# Deterministic Extractor Skill

Generates and runs OS-aware scripts to extract structural codebase data without AI analysis.

## When to Use This Skill

- Phase 0: File inventory enumeration
- Phase 1: Role tag assignment and clustering (pattern-based)
- Phase 2: Signature extraction (class, function, import detection via regex)
- Phase 3: Dependency graph construction (from Phase 2 import data)
- Phase 4a: LOC counting and cyclomatic complexity calculation
- Any time structural data can be extracted deterministically rather than by AI

## Unitary Function

**ONE RESPONSIBILITY:** Generate and execute deterministic extraction scripts, write structural JSON artifacts.

**NOT RESPONSIBLE FOR:**
- Semantic analysis, summaries, or business logic (those are AI sub-agent tasks)
- Calibrating regex patterns (see vm-script-calibrator)
- Report generation or HTML output (see 3i-RE-Report-Generator)

## Input

```json
{
  "repo_path": "/absolute/path/to/repo",
  "output_dir": "docs/codebase-analysis",
  "script_config_path": "docs/codebase-analysis/intermediate/script-config.json",
  "phase": "phase0|phase1|phase2|phase3|phase4a",
  "host_os": "windows|linux|macos"
}
```

## Script Templates

### Phase 0: File Inventory

**Purpose:** Enumerate every file in the repository, classify by extension, exclude vendor/build directories.

**PowerShell (Windows):**
```powershell
# Phase 0: File Inventory
param(
    [string]$RepoPath,
    [string]$OutputPath,
    [string[]]$ExcludeDirs = @("node_modules", ".git", "dist", "build", "__pycache__", ".venv", "venv", "vendor", "obj", "bin", "target")
)

$excludePattern = ($ExcludeDirs | ForEach-Object { [regex]::Escape($_) }) -join '|'

$files = Get-ChildItem -Path $RepoPath -Recurse -File |
    Where-Object { $_.FullName -notmatch $excludePattern } |
    Select-Object @{N='path'; E={ $_.FullName.Replace($RepoPath, '').TrimStart('\','/').Replace('\','/') }},
                  @{N='extension'; E={ $_.Extension.ToLower() }},
                  @{N='size_bytes'; E={ $_.Length }}

# Language mapping
$langMap = @{
    '.py'='python'; '.js'='javascript'; '.jsx'='javascript'; '.ts'='typescript'; '.tsx'='typescript';
    '.java'='java'; '.cs'='csharp'; '.go'='go'; '.rs'='rust'; '.rb'='ruby'; '.php'='php';
    '.css'='css'; '.scss'='scss'; '.less'='less'; '.html'='html'; '.md'='markdown';
    '.json'='json'; '.yaml'='yaml'; '.yml'='yaml'; '.xml'='xml'; '.sql'='sql';
    '.sh'='shell'; '.ps1'='powershell'; '.bat'='batch'
}

$enriched = $files | ForEach-Object {
    $lang = $langMap[$_.extension]
    if (-not $lang) { $lang = 'other' }
    [PSCustomObject]@{
        path = $_.path
        extension = $_.extension
        language = $lang
        size_bytes = $_.size_bytes
    }
}

$counts = $enriched | Group-Object language | ForEach-Object {
    @{ $_.Name = $_.Count }
} | ForEach-Object { $_ }

$inventory = @{
    generated_at = (Get-Date -Format 'o')
    repo_path = $RepoPath
    host_os = 'windows'
    script_type = 'powershell