Security
Comprehensive security framework for LobsterAI providing audit logging, role-based access control (RBAC), input validation, output sanitization, code scanning, and dependency vulnerability detection.
Core Features
Audit Logger
- Records all skill executions with timestamps, user context, and input/output
- JSON-structured logs for easy analysis
- Automatic log rotation (90 days retention)
- Encrypted storage support
Authorizer (RBAC)
- Role-based access control configuration
- Fine-grained permission management
- JSON-based role definitions
- Session validation
Input Validator
- Path traversal prevention
- Dangerous command detection (rm, del, eval, etc.)
- Cron expression validation
- Working directory restrictions
Output Sanitizer
- Automatic redaction of sensitive data (passwords, API keys, tokens)
- Error message sanitization
- Safe error formatting for user display
Code Scanner
- Static code analysis for common vulnerabilities
- Pattern-based malicious code detection
- Configurable scan rules
Dependency Scanner
- NPM vulnerability scanning (npm audit)
- Python package vulnerability detection (pip-audit)
- Automated dependency checking
Usage
All security features are available for import by other skills:
from security.audit_logger import audit_log_skill_start, audit_log_skill_end
from security.authorizer import Authorizer
from security.input_validator import InputValidator, ValidationError
from security.output_sanitizer import sanitize_text, create_safe_error
from security.code_scanner import CodeScanner
from security.dependency_scanner import DependencyScanner
Configuration
Copy rbac_config.example.json to rbac_config.json and customize roles and permissions.
Scan Scope
The code scanner and dependency scanner are designed to scan all skills in the SKILLs directory when invoked explicitly (e.g., python -m security.code_scanner --skill all). This allows comprehensive security assessment across your entire LobsterAI installation.
Privacy Note: Scanning all skills grants this module read access to all skill code and dependencies. This is intentional for a security audit tool, but users should be aware of the broad read scope. Ensure you trust the skill source before enabling full-system scanning.
To limit scanning to specific skills, invoke with explicit skill IDs:
python -m security.code_scanner --skill web-search --skill scheduled-task
Environment Variables
This skill requires the following environment variables to function correctly:
| Variable | Description | Required | Default |
|---|---|---|---|
LOBSTERAI_HOME |
Base directory for LobsterAI data and logs | Yes (if not using default) | ${APPDATA}/LobsterAI (Windows) or ${HOME}/.config/LobsterAI (Linux/macOS) |
LOBSTERAI_AUDIT_SECRET |
HMAC secret for audit log signature (optional) | No | - |
LOBSTERAI_USER_ID |
Current user identifier for audit trails | No | anonymous |
SKILLS_ROOT |
Path to the SKILLs root directory | No (auto-detected) | Parent directory of the current skill |
Note: Ensure LOBSTERAI_HOME/logs/security/ exists and is writable.
Deployment
See DEPLOYMENT.md for detailed deployment instructions, security hardening checklist, and incident response procedures.
Testing
Run python tests.py to execute the test suite.
Integration
This module integrates seamlessly with LobsterAI's skill execution pipeline, providing:
- Pre-execution validation
- Runtime monitoring
- Post-execution sanitization
- Comprehensive audit trails
Security Maturity
⭐⭐⭐☆☆ (3/10) - Active development
We are continuously improving our security posture. See SECURITY.md for the complete security architecture and best practices.