Add AgentShield — open-source AI agent firewall and audit log

[brigen]

Description

Adding AgentShield to the Agent Firewalls & Gateways (Runtime Protection) section.

What is AgentShield?

AgentShield is an open-source security layer that sits between AI agents and their tools as a policy-enforcing proxy. It provides:

• 3 interception layers: MCP tool calls, CLI commands, and HTTP requests
• Policy engine: Rules-based allow/deny/warn with rate limiting and regex argument matching
• Audit logging: Structured JSON logs with automatic secret redaction (AWS keys, GitHub tokens, API creds)
• Web dashboard: Real-time monitoring at localhost:4040 — call feeds, analytics, anomaly alerts
• HTTP proxy: Blocks exfiltration attempts (pastebin, ngrok), detects secret leaks in outbound requests
• CLI guard: Prevents destructive operations (rm -rf, git push --force)

Installed via npm, integrates with Claude Code via MCP server config.

Checklist

• Tool is directly related to AI agent security (runtime protection)
• Open-source (MIT license)
• Actively maintained (recent commits)
• Follows the required formatting: **[Name](URL)** - Description ending with period.