Sentinel AI — Real-time safety guardrails for Claude Code

[MaxwellCalkin]

Resource Information

Name: Sentinel AI
URL: https://github.com/MaxwellCalkin/sentinel-ai
Category: Security / Safety Tools

Description

Real-time safety scanning for Claude Code — detects prompt injection (12 languages), PII leaks, harmful content, dangerous tool calls, and OWASP code vulnerabilities. Works as a PreToolUse hook, MCP safety proxy, and CLI.

Key features:

• 9 safety scanners with sub-millisecond latency
• MCP safety proxy (sentinel mcp-proxy) wraps any MCP server with safety scanning
• Claude Code attack vector detection (poisoned repos, authority impersonation, API key exfiltration)
• Code vulnerability scanner for generated code (SQL injection, XSS, command injection, hardcoded secrets)
• 530-case benchmark at 100% accuracy, 457 tests

Install: pip install sentinel-guardrails
License: Apache 2.0

[hesreallyhim]

This submission was not made through the required web form. As noted in CONTRIBUTING.md, recommendations must be submitted using the web form.

A cooldown of 7 days has been applied. Please use the web form for your next submission.