Blazing-fast, zero-dependency security scanner for the AI-assisted development era.
Identify high-risk vulnerabilities and "ghost" threats introduced by AI agents even before they reach your CI/CD pipeline.
AI agents are rewriting the world, but they also introduce new attack surfaces. GhostCheck bridges the gap between traditional SAST and AI-native security, ensuring your code remains secure while you move at AI speed.
GhostCheck is built with the AgentCortex philosophy, ensuring that AI-assisted security is built on a foundation of verifiable engineering directives.
- 🔌 Plugin Architecture: Fully decoupled scanners and reporters. Extensible support for custom logic.
- 🛡️ Red Team Hardened: Built-in protection against chaos tests, bypass attempts, local RCE vectors, and directory traversal.
- 📊 Universal Reporters: Native support for
console,json,html,owasp-llm, andsarifoutputs.
- 🚀 Framework Presets: Automated scan strategies for Next.js, Flutter, Django, FastAPI, and Terraform.
- 🛡️ Robust Baseline: Content-hash based fingerprinting (
file:rule:hash). Findings stay suppressed even if line numbers shift. - ⚡ Preset-Aware Performance: Optimized I/O by skipping irrelevant modules based on project type (e.g., ignoring Docker checks in pure Flutter apps).
- 🎯 OWASP LLM Top 10 Report: Industry-first
--format owasp-llmsupport, mapping findings to standardized AI security categories. - 🤖 MCP & AI Supply Chain Audit: Auditing for Model Context Protocol (MCP) configuration to prevent tool poisoning and excessive agency.
- 🔑 AST-Powered Secret Detection: Context-aware parsing for 50+ providers using language-specific AST scanners (Python, JS/TS, Go, Java, Dart).
Recommended for most users:
pip install ghostcheckgit clone https://github.com/KbWen/security-tools.git
cd security-tools
pip install -e .Generate tailored .ghostcheckignore and ghostcheck.toml with Automatic Framework Detection:
ghostcheck init# Scan the entire project for all risks
ghostcheck scan .
# Scan ONLY the files you are about to commit (Blazing Fast)
ghostcheck scan --staged| Feature | Command | Target |
|---|---|---|
| Full Security Scan | ghostcheck scan |
Entire Workspace / Git Diffs |
| Dependency Check | ghostcheck check-deps |
requirements.txt, package.json |
| Secret Detection | ghostcheck check-secrets |
Logs, Source, Docs |
| Rule Audit | ghostcheck check-rules |
.agent/, .cursor/ |
GhostCheck respects professional workflows:
- Custom Exclusions: Use
.ghostcheckignoreto silently bypass safe paths. - Severity Filters: Run scans with targeted focus using
--severity [CRITICAL|HIGH|MEDIUM|LOW]. - Multilingual Support: Define custom safe keywords in
ghostcheck.toml(custom_safe_keywords = ["нельзя"]) to prevent false positives in non-English documentation. - Automation Ready: Export results natively using
--format json,--format html,--format sarifor--format owasp-llmfor seamless compliance reporting.
Developed with ❤️ for the AI community by KbWen.