Security

A static analysis and attack-surface scanner for AI agents that maps an agent's tool access, prompt risks, and MCP integrations from source code, then turns material authority changes into CI/CD pass, review, or block decisions.

• Security
• CI/CD

An open-source safety and control layer for AI agents that blocks unsafe tool calls, pauses for human approval on risky actions, enforces per-agent cost budgets, and keeps an audit trail that can survive a crash.

• AI evals
• Security

An open-source gateway that enforces Model Context Protocol tool-call policy inside a hardware trusted execution environment, so a governed agent process cannot alter or bypass its own policy engine.

• MCP servers
• Security

A curated collection of research papers on GUI agent safety, organized by environment and keyword, covering benchmarks, datasets, attacks, defenses and evaluation frameworks for computer-use and mobile or desktop agents.

• AI evals
• Security
• Papers

A deployment and configuration guide for running Strix, an autonomous AI penetration-testing agent, on Kali Linux, covering reconnaissance, proof-of-concept generation, and multi-provider LLM setup.

• Security
• Tutorials

An autonomous offensive-security agent that plans an engagement, runs reconnaissance, chains exploits and writes a report end to end from inside Claude Code, Gemini CLI, Codex or Cursor.

• Security

A reference collection of Windows commands for system administration, security auditing and incident response, grouped into eight categories such as firewall, logs and persistence checks.

• Security
• Tutorials

An OpenAI-compatible gateway for attested confidential inference that publishes hardware attestation for itself, verifies its upstream model providers before forwarding a prompt, and signs a receipt for every response.

• LLM gateways
• Security

An agent skill that pentests your own running application: it finds a vulnerability, proves it with a real HTTP request, patches it, then re-verifies the fix.

• Agent skills
• Security

A self-hosted identity provider for AI agents that gives each agent its own revocable identity, authority that can only narrow when delegated, and a tamper-evident audit trail it cannot write to itself.

• Security
• Auth

A deterministic, read-only static auditor for GGUF model files that flags chat-template backdoors, SSTI risks, and suspicious tokenizer, config or model-card content without ever rendering templates or loading tensor weights.

• Security

An open, MIT-licensed database of confirmed-malicious IP addresses and disposable email domains, rebuilt every 30 minutes from honeypot sensors and community reports, for use in firewalls, WAFs and signup checks.

• Security
• Databases

An autonomous AI pentesting tool that runs a staged, evidence-backed security audit of web apps and APIs across identity, injection, business-logic and configuration risks, for targets the user is authorized to test.

• Security

A curated index of over 570 publicly disclosed CVEs affecting the Model Context Protocol ecosystem, mapped to the OWASP MCP Top 10 risk categories for MCP servers, SDKs, gateways and clients.

• MCP servers
• Security

A penetration-testing management platform for security researchers, tracking targets, findings and tool output, and building a knowledge graph of relationships between them to visualize attack paths.

• Security

This repository holds the notebooks, exercises and data for a one-day masterclass that teaches engineers and architects to build security and privacy into AI and machine learning products by design rather than adding it afterward.

• Security
• Tutorials

Grantex is an open-source delegated authorization protocol for AI agents: it gives each agent a verifiable identity and scoped, revocable authority to act for a human or organization, with an immutable audit trail, complementing OAuth 2.0 and MCP rather than replacing either.

• Security
• Auth

Ablation is a reverse engineering framework offering the same core disassembly, decompilation and binary analysis capabilities as tools such as Ghidra, IDA Pro and Binary Ninja, that becomes a fully autonomous reverse engineering tool when paired with Claude Code or OpenAI Codex.

• Debugging
• Security

A face and fingerprint authentication system for Linux login, lock screen, sudo and app prompts, using infrared cameras with anti-spoofing checks on by default and TPM-sealed embeddings.

• Security
• Auth

Invarail is a local, self-hosted authority layer for AI agents that separates what an agent is allowed to do from what it can learn to do, sitting in front of interchangeable workers for coding, repeatable procedures and open-ended tasks.

• Agent frameworks
• Security

BlueRock is a lightweight runtime security sensor for Python processes and MCP servers, recording tool calls, session lifecycle events, module imports and subprocess execution as structured logs with no code changes.

• MCP servers
• Security

An ACL paper's official code for a logical data-poisoning attack on GraphRAG systems, which swaps entities of the same type to corrupt a corpus's implicit reasoning graph while keeping the poisoned text locally plausible.

• Security
• Papers

A cryptographic trust layer for AI agent actions that signs every tool call as a hash-chained, offline-verifiable receipt, so an agent's actions and the authorization behind them can be proven rather than only logged.

• Agent frameworks
• Security

An MCP server that lets an AI assistant explore Aura's fully homomorphic encryption tools, running numeric computation on encrypted data without exposing the underlying values.

• MCP servers
• Security