#indirect prompt injection
122 related articles
Smart Proxy: The Key Architecture for …
·3 min
Smart Proxy: The Key Architecture for Running AI Coding Agents Safely and Autonomously
Too much AI permission is risky; too little kills productivity. A smart proxy acts as a controllable middleware layer — intercepting, auditing, and policy-gating AI agent actions in tools like Claude Code and Cursor for safe autonomous operation.
Read more →
Codex Encrypts Sub-Agent Prompts: The …
·2 min
Codex Encrypts Sub-Agent Prompts: The AI Black-Box Trend and What Developers Should Do
OpenAI Codex is encrypting sub-agent prompts, sparking debate about AI transparency. We analyze the motivations, community concerns, and strategies for developers navigating the black-box trend.
Read more →
Ghostcommit Attack Explained: How Mali…
·2 min
Ghostcommit Attack Explained: How Malicious PNGs Bypass AI Code Review
Ghostcommit is a new supply chain attack targeting AI coding tools that hides malicious instructions in PNG pixels, bypassing automated code review to steal .env secrets via indirect prompt injection.
Read more →
·4 min
From 50% to 92%: A Practical Methodology for Teaching AI Agents to Master Spreadsheets
An engineering team spent four months raising an AI coding agent's spreadsheet accuracy from 50% to 92%. Deep dive into REPL architecture, validation loops, and domain knowledge injection.
Read more →
Fable Extended to July 19: Challenges …
·2 min
Fable Extended to July 19: Challenges and Lessons from AI Interactive Narrative Products
An in-depth look at what Fable's delay reveals about AI narrative products: technical bottlenecks in LLM coherence, cost sustainability, and content safety challenges.
Read more →
·6 min
GPT-5.6 vs Grok 4.5 on the Same Day: A Packed Week of Flagship AI Model Launches
OpenAI's GPT-5.6 launches with Sawa, Terra, and Luna sub-models the same day as Musk's Grok 4.5, while Anthropic, Meta, and NVIDIA make their moves. A packed week of flagship AI launches.
Read more →
·6 min
AI Agent Security in Practice: 7 Attack Techniques and a Five-Layer Defense System
An in-depth breakdown of the 7 major attack techniques against AI agents (prompt injection, data poisoning, image attacks, etc.) and a five-layer defense system, with real cases from Doubao and DeepSeek.
Read more →
·4 min
Eight Text-to-Image Models Compared: Which Excels at Realism, Ghibli, and Anime Styles
A comprehensive comparison of eight mainstream text-to-image models including Krea2, Flux2, and Qwen Image, covering realistic portraits, Ghibli, 3D anime, and Japanese anime styles.
Read more →
·4 min
AMD Ryzen AI Halo Hands-On: Can x86 Architecture Challenge NVIDIA's DGX Spark?
In-depth review of the AMD Ryzen AI Halo mini AI box: powered by the Ryzen AI Max Plus 395 (Strix Halo) chip with 128GB unified memory, priced at $4,000. Compared against NVIDIA's DGX Spark across token generation, prefill speed, and x86 advantages.
Read more →
·4 min
Surge in Claude Mythos Preview Vulnerabilities: How Can We Ensure LLM Security?
After the release of Claude Mythos Preview, critical security vulnerabilities surged, raising widespread concern. This article analyzes the tension between rapid iteration and security, explores LLM attack surface challenges, and offers practical defense strategies.
Read more →
·7 min
OpenClaw vs Hermes: An In-Depth Comparison and Selection Guide for AI Agent Tools
An in-depth comparison of OpenClaw and Hermes Agent, covering skill management, memory mechanisms, security, and gateway configuration to help you find the right AI agent solution.
Read more →
·4 min
How Background AI Agents Can Automatically Maintain All Your Repositories on a Schedule
OpenInspect's Multi-Repo Automations lets AI coding agents maintain up to 10 repositories on a schedule simultaneously — isolated sessions, independent PRs, and fault-tolerant execution for security sweeps, dependency upgrades, and framework migrations.
Read more →
·5 min
Prismata Analysis: How to Defend Web Agents Against Cross-Site Prompt Injection Attacks
Cross-site prompt injection is becoming the trickiest security threat for Web agents. This article analyzes the Prismata project's 'confining defense' approach—controlling injection's blast radius via context isolation, permission boundaries, and trust grading.
Read more →
·5 min
ECC: An Agent Enhancement Framework for AI Coding Assistants — Five Core Capabilities Explained
ECC is an agent optimization framework for AI coding assistants like Claude Code, Cursor, and Codex, enhancing them with skills, memory, security, and research-first development capabilities.
Read more →
·3 min
AI Compliance Auditing: How to Build an Adversarial Testing Evidence System That Withstands Scrutiny
AI compliance is shifting from document storage to generating credible adversarial testing evidence. Learn how TRAIGA, NIST RMF, and ISO 42001 shape audit-grade red team testing requirements.
Read more →
·7 min
Turning Any AI Agent into an Orchestration Hub: A Deep Dive into Multi-Agent Collaboration Architecture
An in-depth analysis of the "any Agent as an orchestrator" design philosophy, exploring the technical implementation of multi-Agent collaboration, context management, and workflow automation.
Read more →
·4 min
AI Coding Tool Security Governance: How Policy Enforcement Guardrails Ensure Enterprise Compliance
As Claude Code, Cursor, and Codex enter enterprises at scale, policy enforcement guardrails are becoming critical. A deep dive into AI coding tool security governance, cross-platform compliance, and enforcement mechanisms.
Read more →
·5 min
Security Swarm Evaluation in Practice: Validating AI Security Tools with Real Vulnerabilities
A deep dive into Security Swarm's evaluation methodology: building test sets from real, recent vulnerabilities to avoid training data contamination and validate its ability to find more bugs at lower cost.
Read more →
LLM Security Benchmarking: Current Sta…
·4 min
LLM Security Benchmarking: Current State, Challenges, and Practical Guide
Why is it so hard to establish unified LLM security benchmarks? This article analyzes core challenges in LLM security evaluation—covering jailbreaks, prompt injection, red teaming, and more—with practical strategies for developers.
Read more →
·5 min
The Complete Guide to Claude Skills: Configure Once, Reuse Your Workflow Forever
An in-depth guide to Claude Skills: from built-in skills and plugins to building your own custom skills and continuous refinement. Learn to end repetitive instructions and build an automated, consistent AI workflow.
Read more →