Tag
#security
16 open-source projects filed under this tag.
agent-vault
Agent Vault is an open-source credential proxy by Infisical that prevents AI agents from leaking secrets — brokered access instead of direct credentials.
anthropic-defending-code-reference-harness
Anthropic's open-source autonomous vulnerability discovery framework for Claude — scan, triage, and patch security bugs with AI agents.
astrid
Astrid is a Rust microkernel OS that treats AI agents like Linux treats processes — WASM sandboxing, capsule architecture, capability tokens, and cryptographic audit trails.
bumblebee
Bumblebee is a Go-based supply-chain scanner from Perplexity AI that inventories npm, pip, Go, and MCP packages across developer machines to detect exposure to compromised dependencies.
crabtrap
Open-source HTTP proxy by Brex that uses LLM-as-a-judge to secure AI agents in production. Static rules + AI policy evaluation with full audit trail.
cubesandbox
CubeSandbox is a high-performance, hardware-isolated sandbox for AI agents — sub-60ms cold starts, under 5MB memory overhead, and E2B SDK compatibility.
cupcake
Cupcake is a Rust-based policy enforcement layer for AI coding agents using OPA/Rego. Deterministic guardrails for Claude Code, Cursor, and OpenCode without consuming model context.
deepsec
Deepsec is an AI-powered vulnerability scanner from Vercel Labs that uses coding agents to find hard-to-find security issues in large codebases.
claude-bughunter
Claude-BugHunter is a 71-skill bundle for Claude Code that turns your AI agent into a senior bug-hunting researcher with 681 disclosed-report patterns across 24 vulnerability classes.
audit
Evilsocket/audit is an 8-stage vulnerability-discovery agent that uses multiple AI agents to find real security bugs in your codebase — built on Claude Code Agent SDK.
microsoft-agent-governance-toolkit
Microsoft's open-source toolkit for governing autonomous AI agents — policy enforcement, zero-trust identity, execution sandboxing, and audit logging. Covers 10/10 OWASP Agentic Top 10.
microsoft-mxc
Microsoft MXC is a sandboxed code execution system with a TypeScript SDK for safely running untrusted code, plugins, and AI agent output across Windows, Linux, and macOS.
nvidia-openshell
NVIDIA OpenShell is a sandboxed runtime for autonomous AI agents with declarative YAML policies, MicroVM isolation, and network-level security controls.
osiris
Osiris is an open-source real-time OSINT dashboard built with Next.js and MapLibre GL — a Palantir alternative with 16 intelligence layers, RECON toolkit, and GPU-accelerated map rendering.
shannon
Autonomous AI penetration tester for web applications and APIs that finds real vulnerabilities with proof-of-concept exploits before they hit production
vigils
Vigils is a local-first control plane for AI agents — audit logs, secret redaction, sandboxed execution, and human-in-the-loop approvals. Rust + Tauri + Chrome MV3.