v2.0.4

sinewaveai/agent-security-scanner-mcpv2.0.4Feb 8, 2026by sinewaveai

AI Summary

Initial open-source release introducing a real-time security scanner for AI coding agents, providing protection against prompt injection and package hallucinations across multiple ecosystems.

Key Highlights

Prompt Injection Firewall with 56 attack patterns
Package Hallucination Detection covering 4.3M+ packages
Vulnerability Scanner with 359 Semgrep-aligned rules
AST-based Analysis with optional tree-sitter support

New Features

Prompt Injection Firewall
Package Hallucination Detection
Vulnerability Scanner
AST-based Analysis
Support for Claude Code, Cursor, Windsurf, and other MCP clients

Full Release Notes

## agent-security-scanner-mcp v2.0.4

🎉 **Initial open source release!**

Security scanner MCP server for AI coding agents, providing real-time protection against:

### Features
- **Prompt Injection Firewall** - 56 attack patterns detecting exfiltration, backdoors, and jailbreaks
- **Package Hallucination Detection** - 4.3M+ packages across 7 ecosystems (npm, PyPI, RubyGems, crates.io, Dart, Perl, Raku)
- **Vulnerability Scanner** - 359 Semgrep-aligned rules with auto-fix templates
- **AST-based Analysis** - Optional tree-sitter support for enhanced detection

### Supported Clients
Claude Code, Cursor, Windsurf, Cline, Kilo Code, OpenCode, Cody, Zed, and any MCP-compatible client.

### Quick Start
```bash
npm install -g agent-security-scanner-mcp
npx agent-security-scanner-mcp init claude-code
```

See [README](https://github.com/sinewaveai/agent-security-scanner-mcp#readme) for full documentation.