DNA Extractor Skill

Extract the "DNA" of any codebase - conventions, patterns, architecture, and tribal knowledge.

Usage

/dna-extractor <path-or-url> [options]

Arguments

• •path-or-url: Either a local filesystem path OR a GitHub URL
  • •Local: /path/to/repo or . or ~/projects/myapp
  • •GitHub: https://github.com/owner/repo.git or https://github.com/owner/repo

Options

• •--level=<level>: Extraction depth (default: standard)
  • •snapshot: ~2 min - Languages, frameworks, structure only
  • •skeleton: ~10 min - + Entry points, core entities
  • •standard: ~30 min - Full extraction
  • •comprehensive: ~2 hrs - + Negative space, git history
• •--output=<file>: Output file path (default: DNA.md in target directory)

Examples

# Extract from local directory
/dna-extractor .
/dna-extractor /path/to/my/project
/dna-extractor ~/code/myapp --level=skeleton

# Extract from GitHub URL
/dna-extractor https://github.com/owner/repo.git
/dna-extractor https://github.com/owner/repo --level=snapshot
/dna-extractor https://github.com/owner/repo.git --output=./docs/DNA.md

Skill Implementation

When this skill is invoked, follow these steps:

Step 1: Parse Arguments

Arguments: {{args}}

Parse the input to extract:

1. •target: The path or URL (first non-flag argument, defaults to ".")
2. •level: Extraction level from --level= flag (defaults to "standard")
3. •output: Output path from --output= flag (defaults to "DNA.md" in target)

Step 2: Determine Input Type

Check if target is a URL or local path:

URL Detection:

• •Starts with https://github.com/
• •Starts with git@github.com:
• •Starts with http:// or https:// and contains github

Local Path:

• •Everything else (relative or absolute filesystem path)

Step 3: Handle GitHub URLs

If the target is a GitHub URL:

1. •

   Normalize the URL:

   • •Remove trailing .git if present for display
   • •Ensure .git suffix for cloning
   • •Extract repo name from URL (last path segment)
2. •

   Clone to temporary directory:

   bash

   TEMP_DIR=$(mktemp -d)
   REPO_NAME=$(basename "$URL" .git)
   CLONE_PATH="$TEMP_DIR/$REPO_NAME"
   git clone --depth 1 "$URL" "$CLONE_PATH"
   

3. •

   Set working path:

   • •Use $CLONE_PATH as the extraction target
   • •Remember to clean up $TEMP_DIR after extraction
4. •

   Error handling:

   • •If clone fails, report error with URL and git output
   • •Check for common issues: invalid URL, private repo, network error

Step 4: Handle Local Paths

If the target is a local path:

1. •

   Resolve the path:

   • •Expand ~ to home directory
   • •Convert relative to absolute path
   • •Verify directory exists
2. •

   Validate:

   bash

   if [[ ! -d "$TARGET_PATH" ]]; then
     echo "Error: Directory not found: $TARGET_PATH"
     exit 1
   fi
   

3. •

   Set working path:

   • •Use the resolved path as extraction target

Step 5: Run Extraction

Execute the DNA extraction process on the working path:

1. •

   Phase 1 - Reconnaissance (Scouts): Run these scouts in parallel:

   • •Structure Scout: Languages, frameworks, directory layout
   • •Config Scout: Configuration files, environment variables
   • •Entry Point Scout: Routes, CLI commands, background jobs
   • •Schema Scout: Database schemas, API specs, message formats
2. •

   Phase 2 - Deep Analysis (Specialists): Based on Phase 1 findings, run relevant specialists:

   • •Domain Modeler: Entities, relationships, state machines
   • •API Extractor: Endpoints, request/response schemas (if APIs found)
   • •Test Analyst: Test coverage, behavioral specifications
   • •Security Analyst: Auth mechanisms, trust boundaries
   • •Convention Extractor: Naming patterns, code style
   • •Infrastructure Analyst: Deployment, CI/CD, monitoring
   • •Negative Space Detector: Expected-but-missing features (comprehensive only)
3. •

   Phase 3 - Synthesis:

   • •Conflict Resolver: Reconcile contradictory information
   • •Confidence Scorer: Assign confidence levels
   • •DNA Renderer: Generate final DNA.md

Step 6: Write Output

1. •

   Determine output path:

   • •If --output specified: Use that path
   • •If GitHub URL: Write to current directory as {repo-name}-DNA.md
   • •If local path: Write to DNA.md in the target directory
2. •

   Write the file:

   • •Generate DNA.md using the DNA template
   • •Include extraction metadata (timestamp, confidence, level)

Step 7: Cleanup

If a temporary directory was created for a GitHub clone:

rm -rf "$TEMP_DIR"

Report completion with:

• •Output file location
• •Overall confidence score
• •Extraction level used
• •Time taken (if significant)

Output Format

The extraction produces a DNA.md file following this structure:

# {Project Name} DNA

> Extracted: {timestamp}
> Confidence: {overall_confidence}%
> Level: {extraction_level}

---

## Identity
- Name, description, purpose, primary language, project type

## Domain Model
- Entities, relationships, state machines, invariants

## Capabilities
- Features, API endpoints, CLI commands, events

## Architecture
- Directory structure, layers, patterns, data flow

## Stack
- Languages, frameworks, databases, dependencies

## Conventions
- Naming patterns, code style, documentation practices

## Constraints
- Security considerations, performance requirements

## Operations
- Build, test, deploy procedures

## Uncertainties
- Items with low confidence or requiring clarification

Error Handling

Level-Specific Behavior

Snapshot (~2 min)

• •Structure Scout only
• •Languages, frameworks, basic structure
• •No deep analysis phases

Skeleton (~10 min)

• •Phase 1 scouts (parallel)
• •Entry points and core entities only from Phase 2
• •Basic synthesis

Standard (~30 min, default)

• •Full Phase 1 and Phase 2
• •Complete synthesis
• •All sections populated

Comprehensive (~2 hrs)

• •Full extraction plus:
• •Git history analysis (churn, ownership)
• •Negative space detection
• •Cross-reference validation