# Security Policy
## Supported Versions
We release patches for security vulnerabilities. Currently supported versions:
| Version | Supported |
| ------- | ------------------ |
| 1.0.x | :white_check_mark: |
## Reporting a Vulnerability
The AGI-MCP team takes security seriously. We appreciate your efforts to responsibly disclose your findings.
### How to Report
**Do not** report security vulnerabilities through public GitHub issues.
Instead, please report them via:
1. **GitHub Security Advisories** (Preferred)
- Go to the [Security tab](https://github.com/muah1987/AGI-MCP/security)
- Click "Report a vulnerability"
- Provide detailed information
2. **Email** (Alternative)
- Create a private issue with label "security"
- Include detailed information below
### What to Include
Please provide:
- **Type of vulnerability** (e.g., SQL injection, XSS, code execution)
- **Affected component** (file path, function name, class)
- **Impact** (what an attacker could achieve)
- **Steps to reproduce** (detailed PoC if possible)
- **Suggested fix** (if you have one)
- **Your environment** (OS, Node version, etc.)
### Response Timeline
- **Initial Response**: Within 48 hours
- **Status Update**: Within 7 days
- **Fix Timeline**: Depends on severity
- Critical: 7-14 days
- High: 14-30 days
- Medium: 30-60 days
- Low: 60-90 days
### Disclosure Policy
- **Coordinated Disclosure**: We practice responsible disclosure
- **Public Disclosure**: After a fix is released or 90 days (whichever comes first)
- **Credit**: We acknowledge reporters unless they prefer anonymity
## Security Best Practices
### For Users
1. **Keep Updated**
- Use the latest version
- Monitor security advisories
- Update dependencies regularly
2. **Secure Configuration**
- Review hook scripts before use
- Validate subagent configurations
- Use appropriate permission modes
- Don't commit secrets to `.agi-mcp/` configs
3. **Database Security**
- Protect `data/agi-mcp.db` file
- Use file system permissions
- Regular backups
- Don't expose database publicly
4. **Hook Scripts**
- Review all hook commands
- Validate input in hooks
- Use absolute paths
- Set appropriate timeouts
- Never execute untrusted scripts
5. **Subagent Safety**
- Limit tool access appropriately
- Use read-only modes when possible
- Review custom subagent prompts
- Monitor subagent activities
### For Contributors
1. **Code Review**
- All PRs require security review
- Look for common vulnerabilities
- Use automated security scanning
- Test error handling
2. **Dependencies**
- Keep dependencies updated
- Review dependency changes
- Use `npm audit` regularly
- Fix vulnerabilities promptly
3. **Input Validation**
- Validate all user inputs
- Sanitize data before database queries
- Use parameterized queries
- Escape shell commands
4. **Error Handling**
- Don't leak sensitive info in errors
- Log security events
- Fail securely
- Handle edge cases
5. **Authentication & Authorization**
- Validate permissions properly
- Use secure session management
- Implement rate limiting
- Protect sensitive operations
## Known Security Considerations
### Hook System
**Risk**: Hooks execute arbitrary shell commands
**Mitigation**:
- Hooks run with user's permissions (not elevated)
- Validate hook commands before adding
- Review hook configurations
- Use timeout protection
- Sandbox hook execution when possible
### Subagent System
**Risk**: Subagents have isolated but privileged access
**Mitigation**:
- Tool restrictions enforced
- Permission modes respected
- All actions logged
- Input validation required
### Database
**Risk**: SQLite database stores all operations
**Mitigation**:
- Parameterized queries (no SQL injection)
- Local file system only
- No network exposure
- File permission protection
### Thinking Mechanism
**Risk**: Evaluates and filters operations
**Mitigation**:
- Safety checks implemented
- Constraint validation
- Risk assessment
- All decisions logged
## Security Features
### Built-in Protections
1. **Input Validation**
- All tool inputs validated
- Type checking via TypeScript
- Parameter sanitization
2. **Database Security**
- Parameterized queries
- No SQL injection vectors
- Local storage only
3. **Hook Validation**
- Timeout protection
- Exit code handling
- Input/output sanitization
4. **Thinking Mechanism**
- Safety constraint checking
- Risk level assessment
- Dangerous operation detection
### Logging & Monitoring
- All operations logged to database
- Session tracking
- Hook execution logs
- Subagent activity logs
- Error and exception logging
## Vulnerability Disclosure Examples
### Example 1: Command Injection
```
Type: Command Injection
Component: src/hooks/hook-system.ts
Impact: Attacker could execute arbitrary commands
Steps to Reproduce:
1. Create hook with malicious command
2. Trigger hook event
3. Command executes with user permissions
Suggested Fix:
Sanitize command input and use allow-list
```
### Example 2: Path Traversal
```
Type: Path Traversal
Component: src/database/infrastructure.ts
Impact: Could read/write files outside project
Steps to Reproduce:
1. Provide path with ../
2. System accesses parent directories
Suggested Fix:
Validate and normalize all file paths
```
## Security Updates
### How to Stay Informed
- **Watch** this repository for security advisories
- **Check** the [Security tab](https://github.com/muah1987/AGI-MCP/security)
- **Review** CHANGELOG.md for security fixes
- **Subscribe** to release notifications
### Applying Updates
```bash
# Check current version
npm list agi-mcp-server
# Update to latest
npm update agi-mcp-server
# Or reinstall
npm install agi-mcp-server@latest
# Rebuild
npm run build
```
## Bug Bounty
Currently, AGI-MCP does not have a bug bounty program. We appreciate all security reports and will acknowledge contributors in our CHANGELOG and security advisories.
## Contact
For security concerns:
- **GitHub Security**: Use the Security tab
- **Issues**: Create with "security" label (for non-critical)
- **Pull Requests**: Include "[SECURITY]" in title
## Acknowledgments
We thank the following security researchers:
- (List will be updated as reports are received)
## Additional Resources
- [OWASP Top 10](https://owasp.org/www-project-top-ten/)
- [Node.js Security Best Practices](https://nodejs.org/en/docs/guides/security/)
- [npm Security](https://docs.npmjs.com/about-security-advisories)
---
**Last Updated**: 2024
Thank you for helping keep AGI-MCP secure! 🔒
