Skip to content

Crash Analysis

Scripts for driving coding agents to perform crash analysis. The pipeline takes input files regarding the crash and codifies it through structured reasoning.

The agent will first examine the input folders and craft a synthesis.md file with objective facts about the crash. The agent will then create both a thesis.md and antithesis.md which then get reconciled into the vulnerabilities.json output and the report.md.

Tip

For a full end-to-end walkthrough, see the crash analysis example.

Example

Using the resolve-cli: 

resolve crash-analysis codex -i folder_with_crash_files # or:
resolve crash-analysis codex -i folder_with_crash_files -s folder_with_source_code

Tip

Providing any custom instructions you have inside of an AGENTS_MUST_README.md file and passing that inside your source code folder with -s can help give important context or additional instructions to the agent.

Overview

The scripts work by delegating reasoning to coding agents (Claude, Codex, or OpenCode) through carefully structured prompts.

Prerequisites

  • At least one supported coding-agent CLI installed on PATH and authenticated:
  • claude for claude
  • codex for codex
  • opencode for opencode
  • A writable /tmp directory (scripts use temporary workspaces under /tmp).
  • Optional, but helpful: Build/install tooling required by the target project (compiler, package managers, etc.).