Skip to main content

​
Hardpoint

Hardpoint is a static security scanner designed specifically for AI-assisted development environments. It detects the Rules File Backdoor attack - malicious instructions hidden in AI configuration files that are invisible to developers but executed by AI coding assistants.

​
Why Hardpoint?

AI coding assistants like Cursor, Claude Code, and GitHub Copilot read configuration files to understand your project. Attackers can exploit this by:
  • Hiding malicious instructions in comments that AI models still read
  • Embedding command injection in mcp.json server configurations
  • Encoding payloads in Base64 to evade visual inspection
  • Backdooring git hooks to exfiltrate credentials
Hardpoint scans for these threats with 9 detection rules across 2 specialized scanners.

​
Quick Start

# Install
go install github.com/dotsetlabs/hardpoint/cmd/hardpoint@latest

# Scan current directory
hardpoint scan

# Scan specific path
hardpoint scan /path/to/project

​
Detection Rules

RuleSeverityDescription
AI-008CriticalSemantic Hijacking - 80+ obfuscation-resistant patterns
AI-005CriticalCommand Injection in MCP configurations
AI-004MediumEncoded Instructions (Base64 payloads)
GIT-001CriticalMalicious Git Hook Pattern
GIT-002CriticalCredential Exfiltration in Hook
GIT-003MediumNetwork Access in Git Hook
GIT-004HighObfuscated Git Hook Content
GIT-005MediumSuspicious Git Remote URL
GIT-006HighSuspicious Credential Helper

​
Scanners

ScannerTarget FilesThreats Detected
AI.cursorrules, CLAUDE.md, mcp.json, .github/copilot-instructions.mdSemantic hijacking, command injection, encoded content
Git.git/hooks/*, .git/configMalicious hooks, credential exfiltration, suspicious remotes

​
Output Formats

# Human-readable (default)
hardpoint scan

# JSON for tooling
hardpoint scan --output json

# SARIF for GitHub Code Scanning
hardpoint scan --output sarif > results.sarif

​
Key Features

  • Zero network access - All scanning happens locally
  • Semantic analysis - Detects Rules File Backdoor attacks with 80+ patterns
  • Obfuscation resistant - Handles leet speak, homoglyphs, Unicode normalization
  • Trust/verify system - HMAC-SHA256 baselines for AI config integrity
  • CI integration - Exit codes and SARIF output for automated pipelines

​
Next Steps

Installation

Install Hardpoint

Scan Command

Learn scan options

Semantic Hijacking

Understand the flagship detection