Ratnadeep Dey Roy·5h agoThe “Out of Scope” Bugs: How to Find, Escalate, and Fix 31 Commonly Rejected Bug Bounty FindingsWhy your report got closed as “Informative”, and how to turn low-impact findings into valid ones without breaking program rules.
Ratnadeep Dey Roy·16h agoThe OWASP Top 10 for Agentic Applications: Every Risk Explained With a FixOWASP has released its first Top 10 for AI agents. Here is each risk in simple terms, a real-world example, and the first control to put in…
Ratnadeep Dey Roy·1d agoAgentic Browser Hijacking: When a Calendar Invite Takes Over Your AIAI browsers read the web and act for you. That makes every page, email and invite a possible attacker. Here is how zero-click hijacking…
Ratnadeep Dey Roy·2d agoA2A vs MCP: Two Agent Protocols, Two Attack SurfacesMCP connects agents to tools. A2A connects agents to other agents. Here is how each one works, where each one breaks, and how to secure…
Ratnadeep Dey Roy·3d agoThe NSA’s MCP Security Guidance: What It Says and What to Do NextIn May 2026, the NSA published its first security guidance for the Model Context Protocol. Here are the key risks, the ten controls, and a…
Ratnadeep Dey Roy·4d agoMCP Tool Poisoning: When the Tool Description Is the AttackYour AI agent reads every tool description on every step. Attackers know this. Here is how tool poisoning, rug pulls and tool shadowing…
Ratnadeep Dey Roy·5d agoSmall Language Models vs Frontier Models for Security: When Is Small Enough?Frontier models are smart but expensive and far away. Small models are fast, cheap and private. Here is how security teams should split the…
Ratnadeep Dey Roy·6d agoAgent Memory Architecture: How AI Agents Remember, Forget, and Get FooledMemory turns a chatbot into a colleague. It also gives attackers a place to hide. Here is how agent memory works, layer by layer, and how…
Ratnadeep Dey Roy·Nov 8Agent Skills vs MCP Servers vs Plugins: Three Ways to Extend an AI AgentSkills teach. MCP servers connect. Plugins bundle. Here is what each one does, where each one breaks, and how to review them before you…
Ratnadeep Dey Roy·Nov 7LLM Wiki vs Vector Database: Does Your AI Still Need Embeddings?Karpathy’s LLM Wiki lets the model write and maintain its own knowledge base. Here is when it beats a vector database, when it fails, and…
