🛡️ Cerberus — Universal Security Skill Book & AI Guardrail Engine

Developed by Araskova Labs
The Autonomous AppSec Guardrail & Security Skill Specification Engine for AI Coding Agents and Enterprise DevSecOps.

🌐 Official Product Page • 📚 Documentation • 📦 NPM Package

👁️ Executive Overview

Cerberus is an enterprise-grade security framework engineered by Araskova Labs to transform how AI Coding Assistants (Claude, Antigravity, Cursor, and custom LLM agents) interact with application security.

Instead of relying on black-box security tools or unconstrained AI prompts, Cerberus introduces the Universal Security Skill Book Specification:

┌─────────────────────────┐     ┌─────────────────────────┐     ┌─────────────────────────┐
│  Query Skill Catalog    │ ──> │   Policy Guardrail Gate │ ──> │   Controlled Execution  │
│  (14 Governed Skills)   │     │ (Strict Risk Level Evaluator) │  │  (Native Scanners + LLM)│
└─────────────────────────┘     └─────────────────────────┘     └─────────────────────────┘
                                             │
                                             ▼
                                ┌─────────────────────────┐
                                │ Audit Evidence Logger   │
                                │ (.cerberus/skill_audit) │
                                └─────────────────────────┘

Key Pillars & Enterprise Capabilities

• Universal Security Skill Book Specification: Standardized JSON manifests (skills/) defining security actions across Reconnaissance, Code Auditing, Vulnerability Testing, Exploit Validation, Malware Hunting, and AI Remediation.
• Deterministic Policy Guardrail Gateways: Every security action evaluates against strict policy risk levels (Passive, ActiveSafe, Intrusive, ExploitValidation, Forbidden) prior to execution, preventing destructive operations.
• Native Security Scanners: Includes high-performance native Rust engines for OSV dependency vulnerability lookup, cloud infrastructure auditing (AWS/Terraform IAM & Security Groups), secret grep, prompt-injection detection, and heuristic malware/backdoor threat hunting.
• AI Auto-Remediation & Code Review: Integrates directly with LLM providers (cerberus-llm) to generate Git patch files (.patch), rewrite overly permissive IAM policies to least-privilege, synthesize edge WAF rules, and provide interactive line-by-line AI security code reviews (cerberus review --diff).
• Native Model Context Protocol (MCP) Server: Implements stdio MCP server support, allowing AI agents to query and execute Cerberus skills directly as native tools.
• Automated Penetration Test Reporting: Compiles execution evidence logs into professional Markdown audit reports (cerberus report generate).

⚡ Quick Start & Installation

Option A: Global Installation via NPM (Recommended)

Install @araskova/cerberus globally from NPM:

npm install -g @araskova/cerberus

Once installed, use cerberus anywhere in your terminal:

# View the Cyberpunk Terminal Boot UI
cerberus banner

# List all 14 governed security skills
cerberus skill list

# Conduct an AI Security Review on unstaged git changes
cerberus review --diff

Option B: Local Project Installation

Add Cerberus as a dev dependency to your project:

npm install --save-dev @araskova/cerberus

Run via npx:

npx cerberus skill list

Option C: Build from Source (Cargo)

Requirements: Rust 1.75+ and Node.js 18+.

git clone https://github.com/ARASKOVA-labs/Cerberus.git
cd Cerberus
cargo install --path crates/cerberus-cli

🛠️ Complete CLI Command Reference

| Command | Description | Usage Example | | :--- | :--- | :--- | | cerberus banner | Displays the interactive cyberpunk terminal boot animation & status | cerberus banner | | cerberus doctor | Checks runtime health, loaded skill counts, and policy engines | cerberus doctor | | cerberus skill list | Lists all 14 available security skills with category and action IDs | cerberus skill list | | cerberus skill show <id> | Inspects the full JSON specification for a specific skill | cerberus skill show secret-scanner | | cerberus skill run | Governs and executes a specific skill action | cerberus skill run --skill cloud-config-audit --action scan-configs --args . | | cerberus skill validate <path>| Validates a custom skill JSON file schema | cerberus skill validate custom-skill.json | | cerberus review | Interactive AI Security Reviewer on file or unstaged git diff | cerberus review --diff | | cerberus report generate | Compiles .cerberus/skill_audit.json into cerberus_report.md | cerberus report generate | | cerberus mcp | Starts the Model Context Protocol stdio server for AI agents | cerberus mcp | | cerberus export | Exports the skill book catalog as JSON or Markdown | cerberus export --format json |

📚 Skill Book Catalog (14 Governed Skills)

Cerberus ships with 14 pre-loaded, governed security skills:

• Code Audit: secret-scanner (API keys, RSA keys, JWT tokens)
• Reconnaissance: repo-architecture-map (Tech stack, API route mapping)
• Vulnerability Testing:
  • cloud-config-audit (Terraform/AWS IAM wildcards & 0.0.0.0/0 ingress)
  • dependency-and-secret-audit (Google OSV API CVE lookup for package.json & Cargo.lock)
  • jwt-audit & jwt-security-auditor (alg: none bypasses & weak signatures)
  • llm-threat-scanner (Prompt injection & jailbreak detection)
  • malware-threat-hunt (Reverse shells, cryptominers, eval obfuscation)
  • network-infrastructure-pentest (Port probing)
  • web-misconfig-review (HTTP response security headers)
• Exploit Validation: sqli-verification (SQL injection validation & parameterized queries)
• Remediation & Defense:
  • auto-remediation (AI Git .patch generation)
  • iam-policy-optimizer (Least-privilege cloud IAM policy re-writing)
  • waf-rule-generator (ModSecurity / AWS WAF rule generation)

🔗 Official Links & Resources

• Official Product Page: https://www.araskova.com/products/cerberus
• NPM Package: https://www.npmjs.com/package/@araskova/cerberus
• GitHub Repository: https://github.com/ARASKOVA-labs/Cerberus

🛡️ License

Araskova Source Available & Open Contribution License. Proprietary & Confidential — Araskova Labs. All Rights Reserved.