security

Local-first governance layer for AI systems that monitors prompts and responses to block secrets, PII, and runaway LLM spending before requests leave the machine.

Blue41 is an enterprise risk control platform for AI agents. The site says it monitors agent behavior in production, detects prompt-injection-style incidents and unauthorized activity, and helps teams control sensitive workflows and compliance risk.

A Blue41 case study on how a banking AI assistant could be abused through indirect prompt injection, and what mitigation layers help reduce the risk.

Chrome extension that brings AI coding help directly into the browser for debugging, code explanations, security reviews, and solution generation.

Open-source Python DDoS traffic monitor that prints live traffic stats to stdout and can output JSON for pipelines and tooling.

A survey of eight tools that sit between AI agents and the services they call, covering credential injection, interception, policy enforcement, and mocking/testing. The page explains what each tool is for, where it fits, and what to watch out for.

A weekly security audit of Atlassian’s remote MCP server, focused on OAuth 2.1, RFC 9728 discovery, PKCE, DCR, and token handling. The page documents why the server’s discovery chain fails and includes HTTP evidence links.

A Docker-based sandbox for running AI coding agents with dropped capabilities, no-new-privileges, and no Docker socket access. It wraps the pi terminal coding agent and supports local file edits, persistent auth, and skills mounts.

Experimental observability tooling for Go services that turns vulnerability scanning into live metrics. It combines govulncheck, Prometheus, and OpenTelemetry to show when vulnerable code is actually executing in running services.

Local-first coding agent workbench that runs agents in secretless microVMs, with support for multiple providers, parallel sessions, and shared repository workflows.

A curated dataset of sandboxes and related approaches for running AI coding agents with isolation, network control, or credential mediation.

A Claude Code-based operating system for running executive workflows across research, communications, CRM, content, and operations, with a split between a public engine repo and a private data repo.

A GitHub Copilot port of Anthropic’s vulnerability discovery harness, with Copilot skills for threat modeling, scanning, triage, patching, and an autonomous sandboxed pipeline for C/C++ memory-safety bug discovery and remediation.

A security-focused article about a Git configuration attack path affecting multiple AI coding agents, with concrete reproduction details and vendor response notes. It also introduces grith, described as an OS-level security proxy for AI coding agents.

A public showcase for Mangudai, an external attack surface management and vulnerability management platform for small security teams. The page describes architecture, scan modules, pricing tiers, and the intended SaaS workflow, making it suitable for a directory listing.

Repolog scans a live website and produces a ranked audit covering on-page SEO, Core Web Vitals, security checks, and AI readiness for major AI search and assistant platforms.

A metadata-first trust control plane for authorized security workflows, evidence retention, release trust, and business-flow proof. The repository includes role-based docs, quick-start commands, safety boundaries, and release-trust materials.

ARIA is a source-available security operations platform that combines a voice-driven 3D cockpit, governed autonomous actions, local-model support, and security connectors for cloud and identity telemetry.

Attestor is a TypeScript control plane for high-risk AI-driven operations. It sits between an AI-generated request and the real system action, applying policy, approval, scope, freshness, replay, and evidence checks before returning admit, narrow, review, or block.

A Noma Security research post detailing a prompt-injection flaw in GitHub Agentic Workflows, with attack flow, proof-of-concept evidence, and security recommendations.

Python framework for conversational social engineering detection using a finite state machine. The README says it tracks turn-by-turn signals like flattery density, give/ask ratio, escalation velocity, and scope mismatch, and exposes a simple `Xtra().analyze(turns)` usage example.

Giskard presents an AI red-teaming and continuous evaluation platform focused on finding hallucinations, security issues, and agent vulnerabilities. The page explains how it compares tools for agent-level testing, regression checks, and guardrail workflows.

Docker blog post about a real AI coding agent failure and how Docker Sandboxes aim to contain destructive execution mistakes.

A research note from OpenShell on using formal methods and Z3 to reason about and verify permission changes in long-running AI agent systems.