discovered 03 Aug 2026
agentseal
→ View on GitHubAgentSeal is a comprehensive security toolkit designed for AI agents, providing robust capabilities such as detection of malicious configurations, tracing toxic data flows, and scanning for potential supply chain vulnerabilities across various agents. It features an extensive pipeline for local scanning, real-time monitoring, and auditing of machine configurations without the need for API keys, alongside the ability to test against 225+ adversarial prompts. Notable functionalities include the `guard` command for scanning and assessing the security of agent configurations and the option to create organization-specific policies through custom rule sets.