RelayShieldAdmin (@relayshieldsecurity)

LLMjacking is draining AI budgets faster than most teams notice. Here's how to catch it before the bill does

A leaked AWS key used to mean someone might spin up crypto-mining instances on your account. In 2026, it increasingly means someone routes inference traffic through your Claude, GPT, or Gemini access

Jul 27 · 4 min readrelayshield.hashnode.dev

From Advisory Tool to Mandatory Gate: A LangChain Reference Implementation

Our last post closed with an open question: a commenter had made the point that a Tool an agent may call isn't the same control as a gate the system must pass before a high-impact action, and we said

Jul 22 · 3 min readrelayshield.hashnode.dev

You Asked, We Shipped: 5 New Agent Security Tools, Structured Results, and Two Bugs We Found Along the Way

When we published the first four smolagents.Tool classes for the RelayShield Agentic Attack Surface, we closed with a question: if there's a fifth tool that would actually get used in your agent pipel

Jul 19 · 4 min readrelayshield.hashnode.dev

Giving agent frameworks a threat-intel safety check: RelayShield's smolagents tools

If you're building autonomous agents with smolagents, two of the riskiest surfaces are usually invisible until something goes wrong: the MCP servers your agent connects to, and the prompts it gets fed

Jul 17 · 3 min readrelayshield.hashnode.dev

RelayShield's Agentic Attack Surface Is Now an MCP Server

AI agents get compromised differently than humans do. A typosquatted MCP server, a prompt-injection attack that exfiltrates credentials, an agent framework sitting on an unpatched CVE, a service-accou

Jul 15 · 2 min readrelayshield.hashnode.dev

RelayShieldAdmin

About

Available for

RelayShieldAdmin's blogs

LLMjacking is draining AI budgets faster than most teams notice. Here's how to catch it before the bill does

From Advisory Tool to Mandatory Gate: A LangChain Reference Implementation

You Asked, We Shipped: 5 New Agent Security Tools, Structured Results, and Two Bugs We Found Along the Way

Giving agent frameworks a threat-intel safety check: RelayShield's smolagents tools

RelayShield's Agentic Attack Surface Is Now an MCP Server

RelayShieldAdmin

About

Available for

RelayShieldAdmin's blogs

LLMjacking is draining AI budgets faster than most teams notice. Here's how to catch it before the bill does

From Advisory Tool to Mandatory Gate: A LangChain Reference Implementation

You Asked, We Shipped: 5 New Agent Security Tools, Structured Results, and Two Bugs We Found Along the Way

Giving agent frameworks a threat-intel safety check: RelayShield's smolagents tools

RelayShield's Agentic Attack Surface Is Now an MCP Server

RelayShieldAdmin

About

Available for

RelayShieldAdmin's blogs

LLMjacking is draining AI budgets faster than most teams notice. Here's how to catch it before the bill does

From Advisory Tool to Mandatory Gate: A LangChain Reference Implementation

You Asked, We Shipped: 5 New Agent Security Tools, Structured Results, and Two Bugs We Found Along the Way

Giving agent frameworks a threat-intel safety check: RelayShield's smolagents tools

RelayShield's Agentic Attack Surface Is Now an MCP Server

Search Hashnode

RelayShieldAdmin

About

Available for

RelayShieldAdmin's blogs

LLMjacking is draining AI budgets faster than most teams notice. Here's how to catch it before the bill does

From Advisory Tool to Mandatory Gate: A LangChain Reference Implementation

You Asked, We Shipped: 5 New Agent Security Tools, Structured Results, and Two Bugs We Found Along the Way

Giving agent frameworks a threat-intel safety check: RelayShield's smolagents tools

RelayShield's Agentic Attack Surface Is Now an MCP Server