Cadence is a command-line tool for detecting AI-generated content in Git repositories and websites. The CLI provides powerful analysis capabilities with flexible output formats and extensive configuration options.

What is Cadence?

Cadence analyzes:

Git repositories - Examines commit patterns, metadata, and code changes to detect suspicious commits
Websites - Analyzes page content for patterns common in AI-generated text
Optional AI validation — Uses OpenAI or Anthropic to provide expert analysis on detected content

Available Commands

Command	Purpose
`analyze`	Analyze Git repositories for suspicious commits
`web`	Scan websites for AI-generated content
`webhook`	Run a webhook server for Git platform integration
`config`	Generate and manage configuration files
`version`	Display version and build information

Quick Start

Analyze a Local Repository

Bash

cadence config init          # Create default config
cadence analyze . -o report.json

Analyze a GitHub Repository

Bash

cadence analyze https://github.com/owner/repo -o report.json

Analyze Website Content

Bash

cadence web https://example.com -o report.json --verbose

Start Webhook Server

Bash

cadence webhook --port 8080 --secret "webhook-secret"

Global Flags

These flags work with any command:

Bash

--config string   Path to configuration file
-h, --help        Show command help

Detection Overview

Cadence uses 38 detection strategies to identify suspicious content:

Git Analysis Strategies (18):

Velocity analysis (additions/deletions per minute)
Size analysis (commit line count thresholds)
Timing analysis (commit intervals and patterns)
File dispersion (files modified per commit)
Ratio analysis (addition vs deletion balance)
Commit message analysis (generic message detection)
Naming pattern analysis (variable and function names)
Structural consistency (code organization patterns)
Error handling analysis (missing error handling)
File extension patterns (file types being modified)
Statistical anomalies (deviation from baseline)
Burst pattern analysis (rapid commit clustering)
Timing anomaly detection (unusual commit timing)
Merge commit detection (merge pattern analysis)
Precision analysis (pattern consistency)
Template pattern detection (boilerplate code)
Emoji usage detection (excessive emoji patterns)
Special character detection (unusual character density)

Web Content Analysis Strategies (20):

Overused phrases (common AI filler phrases)
Generic language (excessive business language)
Excessive structure (over-organization with lists/headings)
Perfect grammar (suspiciously uniform sentence lengths)
Boilerplate text (common filler phrases)
Repetitive patterns (repeated sentence structures)
Missing nuance (excessive absolute terms)
Excessive transitions (overuse of connectors)
Uniform sentence length (unnatural consistency)
AI vocabulary (characteristic word choices)
Emoji overuse (excessive emoji in content)
Special characters (unusual character usage)
Missing alt text (images without accessibility)
Semantic HTML issues (improper tag usage)
Accessibility markers (missing ARIA attributes)
Heading hierarchy issues (improper heading structure)
Hardcoded values (inline styles, pixels, colors)
Form issues (inputs missing labels/types/names)
Link text quality (generic or non-descriptive links)
Generic styling (lack of CSS variables, inline style overuse)

Command Details

See CLI Commands for complete reference including all flags, options, and examples.

See Detection Strategies for in-depth documentation on each detection method.

Configuration

Cadence uses YAML configuration files for thresholds and settings. Generate a default config with:

Bash

cadence config init

This creates .cadence.yaml with detection thresholds:

YAML

thresholds:
  # Size-based detection
  suspicious_additions: 500
  suspicious_deletions: 1000
  
  # Velocity-based detection
  max_additions_per_min: 100
  max_deletions_per_min: 500
  
  # Timing-based detection
  min_time_delta_seconds: 60
  
  # File dispersion
  max_files_per_commit: 50
  
  # Ratio-based detection
  max_addition_ratio: 0.95

Output Formats

The analyze command detects output format from file extension:

JSON Format:

Bash

cadence analyze /repo -o report.json

Text Format:

Bash

cadence analyze /repo -o report.txt

The web command uses --json for JSON output, otherwise text:

Bash

cadence web https://example.com --json -o report.json
cadence web https://example.com -o report.txt

[!TIP] The reporter system supports 5 formats (JSON, text, HTML, YAML, BSON). All formats are available via the webhook server API. The analyze CLI currently supports JSON and text.

Common Workflows

Quick repository scan:

Bash

cadence analyze /path/to/repo -o report.json

Analyze GitHub repository:

Bash

cadence analyze https://github.com/owner/repo -o report.json

Website analysis with details:

Bash

cadence web https://example.com --verbose -o report.json

Set up continuous monitoring:

Bash

cadence webhook --port 8080 --secret "my-secret"
# Configure GitHub webhook to point to your server

Next Steps

CLI Commands - Complete reference with all options
Detection Strategies - Learn how each strategy works
Repository Analysis - Real-world analysis examples
Configuration - Advanced settings and customization

CLI Reference