Keyboard shortcuts

Press or to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Overview

Odin Scan processes smart contract code through a proprietary analysis pipeline that combines AI analysis with deterministic static checks and a verification layer. The pipeline is designed to maximize detection accuracy while minimizing false positives.

How It Works

When you submit code for analysis, Odin Scan performs the following high-level steps:

  1. Platform Detection — Odin Scan examines your repository to determine whether you are building on CosmWasm, EVM, or Solana. You can also specify the platform explicitly.

  2. Context Understanding — Odin Scan gathers context about your project (README, compiler version, audit history, trust model) to tailor the analysis to your specific environment.

  3. AI + Static Analysis — Multiple AI models and deterministic static analysis rules analyze your code in parallel. Each analyzer operates independently to maximize coverage. See Multi-Agent AI and Static Analysis for more on each approach.

  4. Verification — Findings are cross-checked, deduplicated, and verified against the source code. False positive reduction runs automatically to ensure the findings that reach your report are accurate and actionable.

  5. Report Generation — Verified findings are formatted into the requested output: JSON, Markdown, or SARIF for GitHub Code Scanning. Each finding includes a title, description, severity, confidence, code location, and remediation guidance.

What You Get

Every scan produces a structured report containing:

  • Findings with severity (Critical, High, Medium, Low, Informational) and confidence levels
  • Code locations pinpointing the exact file and line number
  • Remediation guidance explaining how to fix each issue
  • Proof-of-concept code demonstrating exploit scenarios (when available)
  • Context-aware severity adjusted for your project’s specific environment

See Vulnerability Categories for the full breakdown of how findings are classified.